{"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TDU5ZdYqEqmp6Fm4FpYpP4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BS8847afEmSS5StksWxaMB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Be99Yy7dKz6EkNxJJo3PNa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RcA9EgvD4pjc76A3t1yzvs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9WvZFPL4sVJGFzMFTXo8Jt"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K8A5jNVc8nudAbMGARfiNp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nr5BkrY29k7jAUuCDMWT7n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-25Xbm52FLCFePv3u4WPwVZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PL5V1USEhyTD6LRcj1a6Gz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5VHv2qWxo4tcKPNMV5gAF6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YE2B5CpFMkqcdXTG2t9XUz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X5j4oVpf9GoXSxpGr5vvce"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NgaDsgwUqJchUpYcygZFLR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TKeLqw4uv1fuQDBrrHASe1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HvyEHpqxQ8xPmkRQkUTxkR"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLbCbTD7ye5CLTpJAMArJY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MZKgDERWpmb5DMoppxxqTq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fbaa1J4DrMp6sXNKhMuoQm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MaxUJipd6pqan5Uo4c1auh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73jofKGjdkrWw1buuyEgGZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GQYX4NSCZyDMg8QevThLQg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-46JhfxyrfYjt3eUhWvr2dZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PUBMazGmmGrL81fr24AQTf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S6DkaWmEHFxgPAdojCrWKg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EWRQviVswe3jYbwY1tAcJs"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFHf7p6BC8SgmivChcxne8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDAzGR1XTQ6uHb4Cjkm7CX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hjydw1phYz1Y6Ko8YpDPkY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BF4PvniUySbU2CNvsnw7AR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Aww1Y7muc5phP9ZUP71H3Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MiLd1FBbsXTGkdv5Z2PLDd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DUYBSm4PnjfPrn4gtziEvc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4wLq4LKV8H8NYbXcePdtV7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NQAbv2fdE3JPT5KaJA58ds"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QHn32sawEwyKyZFeyhiGij"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Mo2WRRW1Y2nvPDFbNCg5i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HxHT4hSPNtCp69Vom9eZ1L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EmYjGykHRURtk4ZU6DXjXx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PeRTCdKjwpdCgdDiCzRJ5S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VhTrsHwm1hYv1tDqPcBuH1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bpiwop9cwuxwGcNhxDroqz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KFrfkYpfms2yAQ2NojGUz9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EVJ4jeEPWRcPRi49SYZGkH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9BSZaGHmKPsYYU5Xw7bad4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BMNXnXVDyUr4qtv8EUe4dc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3PcdsR4iXSp2yQwDWD8dYh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ECY55pM63r6CWpvG5u8iKk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q9ecKK7ogv6FLrc2Jx9fxR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XGqBYg25knnw5mrPm4PHu2"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AMiPdZwJEtzDsr8B6qvxS8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HdKXwb6byWx1cPzPyHcRbw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LBbsqHNjgCCvXLDuNYfV3B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ny7yejLrJeB5XAW4sSE8N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CungtyjvBpNR2NGurwaM3X"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hy9UhcgVd15vEQzGyqLeL4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pr9tWj5AbSP6DXtHacHPQ5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KxsuMxgRjNFncmQPdSVNjy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3L94dXPyLTBh6cB33LP28J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RKK7Sado53xas2A7yNDkNv"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7wdx8bahcGzBsuTJGzhMtw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PNuQG2k94yufznpzutXQom"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WziNJ4Ag6Nq5tmmSF18n3M"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2R7wLaFxnnzDwLG7ZuaDbw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3YRfhBGEy5R2vwyHaX3h8M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XtzKVrVLGANmnLkUWy37bt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QAVtYzrhKPPxLyEgEgbSHF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Y3WrRp12SAZBiE1si6eza"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LWRvG1L5hub3Hi1ZJMX3gW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9zJAHuuXCm47nZY93n1Peb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AXTAfLKnTqY9U8PHtjC4eU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1yWNizdJkkx9udb3gVxC2w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CFgExyAQc93RamQFXbEHNq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LHRTpm7eH1HTKv3UxCqFox"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ExZmk3AejM3WG5s7g4JVwW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UkmDv4iwAv9U1Ua2Ap8Wwa"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RnEdMJ5bQvxvDThr1Q1qia"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U5L3sFHC29nymQWoNrWF1f"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7bfE6dLVgHxJLhvFnFbGTj"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFhTgbFFpWN6faqcuMp2nt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ks7tfJVxUXZZjrU6P28wwK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FyGSJdmHhzEQ65h5ToEGkp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jaz1MVwrxa7UDWVPEU5thH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U9Pgcq23zfkwmyXHB7wTE3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4oqiFPASH1Qmo1SUSbhpk2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-USGmDVdc7zEs9quoDYzWQW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3c2r1xNNwUXDcXof9ApS3q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MMuiZsCwUMKbGXr7grQ39b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ZYBGabbxhZY3CrHtoPddF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SbbMTqmMa64B6tXv3dFThF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CuX2RZkBJhYyZpgcE8hbnG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JxCbKWXwsqdrf78Y2KXvxm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SSrZSjo8vQakkWgaaZjvds"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73gSA6BRRTbiQqq57gqEGu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8KVGfZXHAHJnEAJTbKEKuR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9WHgCSQj2v9xPX8Lm3sQST"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7isidrYuriKkfGWJEHCtNY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WjnRjMbRT79eiDWAejdeMH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8tXKwFBZkC2TYXJatpSpTS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TfcqzjmRd41TJEGirggP9w"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PLb6YfeN9GWMLa5AKLKbEH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-57pehBKXfVXxebx5ANwykM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GyvVgNPVZVc7nFZvRoPLwK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9CTYpkHwNkVCdnt9foZPtL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQXdurYKWZQWisWuqgz7PX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFrmpqtRNimxwJVLEPAmjr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VgWPsLVxroH862jJJpyXwt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WD9WZco7twq9S63H78w9YE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LPXVZtA11qikt5QYxv6B8c"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4MoF5sERNUKEABb7gXs9x3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hv1pvoFMNoCbPWfzEVgpG6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1u4eNVuRmZNHtDHhGMqTJT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ELZtEP1jpUujTR8RiNCsDR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6V3CJ8LaTU5T4K128m4tAB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EU1JwM4f6BBZZyNNXmX7Fs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J6RT1Vcu5eBSjjYLCkeMPY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DJijqkANJNjTBxE6GAdkA9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EM1CBDFmcdpAMbLqrPcWJS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UR1FznJfdPL92tB3P4HD1J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W15Ss3KuQkKUfQYPkwmbQF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2t7PjSerPRdrDhQnac1cK6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EkFFHYMaVCLkaVN2tK95wx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SHwuWAty1nsuWucdwYw8jH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6yxFmSAj4XXf5KywLm9RkQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SVACC2g3fWsgnmCf5x6iQn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4bennsfmDJiJYWgxjUbUZT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KhZXgT8PKYNYWtQBv4nvr5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GquhWSNxyggeKMXesF11jY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wG4spGoZUnKCrkxqkJ62R"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XzJM7CJgy2garC8RLhu8g7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fhawx97mjT7Dwy1PJazwqc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KFF5fCM3PZ8tihbfZsUfdq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RKzw97dtZKX8VMvVXwJPR9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQbAKUDnFHaJRMq1d7zdS8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BYKyddAeECZDcKt8FJMXDe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U6gSTcjJSmqmxJRe2FafUb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DjWtqoCEDzrEPvmEtF6WiQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AQfZwjUD7WBsWXXDBChewb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1gL82F1Nh42rxCTriXqyXU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pi93NngTVM5wtMweMMjZFp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LkoiLR3PEGeFq8wPQrUgPi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CKsdYM2LwNsKJ5yvU6APC9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7LkUufbSLQo57vu1m85C13"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NLuinVmjqVdwVDoaoDMyWK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FwzDNmoYFGasKxHoGVD2He"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MZMS4vDN2vCSMrNAUyQbjt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KFpRaE8HFBHyFyZbBjM4cL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GP9sAuKFg2MqBHgJ4nNt6d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HSp5Zv2jAeXsBJ9QTb5qxM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MaKqdJ4w2X58G9sFBRRf3c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BL2WPwReF5vqTXQt8aMX8R"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CwX6ezoBdk2hRFqKMQ48Gv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FqXGbQsoQ37DjEd7Qwwifs"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QhWnrXDZc2CHAFRBJFAetC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ms4Kn7VZfFxr3Sxwv9HGGw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TM55ov9wESh6fFSC4Rf9VM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8VrAwKxMyxYhpYxpdAviQE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DEaQgxLPCzx4NWSoqyc8hx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UFU9KR2zDSy4HEYrEPwViN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3HwtQrvG2BGb8LD7VfWdJd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RFBsYvEpZaj5pd8XXYKrfG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WJb1vD1euu4rxgmmK9G8Dx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L55Lv8fWWqCXSBaJvctrJT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TNpfP6LzDiSpnm6BGCpfwS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8KpEJti2L4D17BnWH76Ui8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DJ9pWJJABQVjUrdnrF2Nxw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SmeCcFcKTbmG1Qk84ZG3A9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RCTiDzsE8pkNQPMNerNBX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Lj4gxZ7qzGn7iDyeTTDf8"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9MNbDXcAqC3ks82HqRpR9j"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GjjyZT2u98q5947ySmMttc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7TCFuQS4MsyeGYR314kbXH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7zJiQ5dPRFamdmHG7riVum"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DjEbnhkVbCbD4gXqvUWJoZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ARs4XapmNgVVQ2sP2LsZ2Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Cs8GEPfiKZnZPCBaKSJdK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PLhF9T35T2J8z7hHYgHnNt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7e4SC4shK3WdVzw118hyBd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CP5MnncCbfXbw4Wx1nxGdE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lt7ofKY4GMss29My27VZm5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FyHNjfaD8K1ZmLbLawVc1L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1aj6skHCBpmBvm7XydYb5w"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1MdcTwNpuNPTseTv56B6yx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B1tt8HrUGHS5b4vaTxsb9Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GUYCz9NfQDAUv8fwqjk3WU"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CLgCeQdeVmr9gQECEVfhPi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1FqcZPiLXk1RrLCSnXsXwY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QpKjL525ywYTivnj98nzDk"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SkrVHopTadPFua6SqFMW23"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Jucv8RRGRqdjwXv4czTg9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sk4eZ7GQB8MZtgCu2M1diR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YGGkdryRHKnnuarSBe78ZD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4opi72c1RmEqfb9NjTQB2h"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SbxrytqDoFei6jEFmN77HP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GyjeA45F5ERB7UPQH2bae1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ih8ofdaSSAGyamc8Wbs3x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wsJrsWPqzGa7zW5nrUeJr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vc2ASCHwuMo6xp4dYjoxxJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M9nFwUGKjrdSS3DhgUc3A3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JBqaXafEUserReD4UuLS58"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VSDRvLdBsP3sQ8JVamhz4V"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3abBCErZWhvKjAtFc2b11"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZrM2BpkcYJCaqPTh7CnHs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FNDfvJ5xRqBBMQHBLW2GG6"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TK8akxND7sYBBgNpp1RJjN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YUSHr7Sj98pq6eL1rEHRcJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LSEmTBLYncMjWbeuVkxmu5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FE7c2v7K3kFdXqWGppqKvx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wnb6YvahUkV28nj1JFrRMy"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DzCGup4yzxFShqwS3LDuUk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDyu6D9r68MFFsmhESo89v"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LW5HPBjygxfHajhajwAony"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4kHJphZKrfqb5yW31sv6Wi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DDUTbXKnPdViHxXBNgtndU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TzhLsJNj6sZUG2MeyFbUNf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KrLE43Ttuq1JxWaPdrncSZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FWtyhoLJgXig5A2WUkkRBf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Azfsn5A5cF5RJghphJJSEP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EzyEG88SbXusYpwPDodzTc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7NM88qqf2c2ZdSgWGCCQyz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9CfErGKUB8da6EKuzzkTbR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JDc49yN1mX6BfUkcyPxgGA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DxyZC3QdMhZCjw1C38x7RR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALVejbt8gDV6H3UseK3kgX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HeR24cfbfDgigb6rVc43Pk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2pZSLnwEw8mgTWkJpGypK4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WAkJCnAnJqR2HV1dtqAoZW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DM6AaQxJEW1TiztubZaCt5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SzMARrN64qdVXw4dcMxKLg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2SfAi9o3KTyzrSVoz9B2ZM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M5W9ZTzrQiXX8ER4tzm3jN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HhVYhoxU4uQ4411J1uivst"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CbJXsK3isJn3sBQqC3SXmy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YKdLhJWskMfcdvLJPi1Xt2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P5kaeSa88PjSWCR5KeAq9R"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QbncpHHuL5ubXTGgGGzupT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CVXRQhVriYimC8NNFLnAeJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3NGb3JAcmsfNh1yvWE5fuJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5NYotqhRqnwpBQrCichR4U"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NnoAjTnK4UZ3if8bixHzXD"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JNqJWvWSvBEMfMWzzg98Xq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JYAQfXCarhsUwgpopDJHBW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TaL5wHbw3Lzvmtqs2MJwdm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GzkEzikgSA8vw2No1mFCn1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NVcWNzTHfvYbrcLRvkL4Mw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-89MidW1rVHsqT9PTZ6r1c2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wt8myLrspEsQxzmgnJ3t57"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CpVZrK742KCGooFzaD2EGM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EeeSbmfi8ZCY5VQTdjfDwb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2TV1m5LzBBBgeGAp7DbR5T"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dk2sLF12cHHyQ3QVoQE5Xo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U6BfG5eLZxW7YYbtjiuGg7"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J6Xmu3iVeckJaUPughh8eC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Np4PGniVco3CJrEDhiKF1b"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4MtF2jfE9c4QAwXR66cSLN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8PEokQHTR3HwutHYe5yqr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-21bS7BWKpRicqPW3p1vZ4A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KyxAjmwnAXS3dLbtKWdXQd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3vck6B5etLKD7mTF5aAfNw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WbunzF4E56tPXDYAwGnQm8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KTGXxv4MRPAgQYqaa4YsiG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HdNDBaiWuU88nKt5sDTEYv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J963SmpetQydtX9TZmHEPp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M1Z1wVtnyUHmEt9urZLRJT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EdVtn1HNcwdFRdccM9AmYe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NmcC48dK5PVCTTUPMWfUbP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8DJpTZLvHuu3uyYo14zEt9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UELrG1bVgUyMGvwNKD8pCn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F91nd5oCGZGfWFwCgLxHQR"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QLW1i9uSE7u77o3FYNjmeK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WgmBDgcomXdhv5axHHqstW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1RaMf1UArH4TaFbhFzBx9Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EMFRiBb5QR94AyCQD6UB4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UnJ7h4sCoY6cc1M2MgTdcU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M9TbMYSmLkGVhgT4Z3cAp2"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBX9su7PoRN2PVDc7ZY1H5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ZKiYyyz7DvhHNwbExbKRo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDNPFTSz6QroJNkbGp7So9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WPkKms6VzYad1nE5Ee235L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AYTBMaRGgAEjoL9sdUpCRa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UrvnCowtmy3Ep67oJUNE3n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Xk4gVKzBzyC6YH139oKP3"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XrpcwTzUS6v27Y3uqw7dLt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CQteafhtEYNimgs8QBoL3a"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1PJQNPBY5RUnPEWGFFA3rw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UMSzXCxYTX6AQif7JU4hV7"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3fpdeD4MCiGKLS4BrVbENH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7K1kWQRZFiGL3Ws3HEASKk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ppxvotf4R8n8mRNijh1Yc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EoLmNMMt836imE1woBqU9x"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8igZzjWLt9wtbXpQAELMUb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2eFoj1EHzPzXAQGm1vWEYj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SxJNCxepeqcezgQHcuUc59"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R1P8DPzfGJK82Xi7tDaU6e"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WycPogWq6bmWqsackcgTZv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RpR4BQMZRdwirPEndK3WJq"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xd1Jc3T5bCR6m3CiKE1Kiq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DVUhUpuemg9j27i5bkFJ3s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EFmmBWTE97vj8iK2dWRTqY"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sj4iRvVT89mTCHvMpgQA4o"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LdPMJ2v6PD6E7g36TXmkU3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7cxtsbe6fya4isC2eeSR4F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JjBNftyq7s23iyyqGMApJC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J23govuJmUUCsavNLyoser"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKyzqS8kSFX2gvunHSsAwN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8sgkCVgSWkHpUkJDBB9o7x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LJkpo52n8822AofdrVkSng"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xc1kZSv1xYh3QcZ4Barorq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2jrVBmVzWC7JxyxUsLSLVM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KfwH1VhqFhL4uDVKBBxvdq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NGM8pU787JBxc4JJwjofAX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WFR8pA4U5rjH9vL7zSQTf1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9fs7joQZnASkRRh3YAkSr2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QnQog72w4BGzBbzngLJo42"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9smHkKof45AV4aDV6hFQjg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CUm5NZoi3jc7YVmkghuKL9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-My7PubdDt8fNJJGkcYGfxy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1MM1g9oM7kvRJ3tYm2Jjqo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GiNadrXtnxB6TwgEcp6hMf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G3bZQ8BMBdSLNrkfwxA7oX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7j3PuCAxwEDAPXVZ2jAr7c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PV2NN4DXSFuj84z1XRxGDN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KGuF4qrb1mqiompscDDHpr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H9uRNZRHKYYXVnzbY6tPCy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B511EmuwyvsSBku5UNedV6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VARbytqFrxvLQFYyHsLTmE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4h83AEWCanVeTrpDNWSfv1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TpvjixmHzQLrYWKNBvGBAt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ST24wzngGtEiDdCbGhoBo4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3U854xTRh9Nu3AYu2xyvRs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PMRBigYrmvcDwmLn5WF4q4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ziKbb4QLhLiRsuQ7tpUur"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FihM5qQzRkUKjXMK69Jgs1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1YUjQyorSsN3GXQ3dLBrBH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JduMo65sHp9hVikyaKXAJC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M4Chc91QNB129KmfnNaTkM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8hWGe683ocWVuMQUKGM5zF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-67JJzWuVfrdmsBbBwyM4Mu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MrbuMLeM8z6LABH2pTrDSq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-54jh985gHc4BHwXAxHaqWN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MY629WGoAadGKwiBgEEULr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jiquh8p5i7hLUtY7b3JxoX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1PkTiPTdmFdsTLQ7CQfquj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K8fRQZeR1oXnckUPUzX5pv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DiLfgJg4RWBANT6BSud9Us"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B6zZhZbwkJBp44v32QL5W1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1M2m1uh6kJb943TiwZictd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XrMnQAZcDxykL4ipiS8Y5F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RXXxXmCkZByMcNqtfEiUhm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T7QvL35TigKq4jyt1o57kJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-75U2eZ1XRBFyL7ihAM61KL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pb1BsMdzGSzy8Ugh4GMHpA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9wSveb2SUotcrsCsspGpg8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4wn5xm8LDTgcDwBMHS7BPQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fa2YujWYKpXXJyH9ZTYnFi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UAqqNYpGmTMaKtNZvttMr1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ERhepFxZcPFekn6rT8aJCF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-24CFLg9NW4U1TPXfHxxkFT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1rzunPLriw99h9LKJ1f9HC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5SnCd4bjHr5rWearJw4RMD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XT2ZTfe8D7C46RQd8dzSCz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Shrgh4zK1gQ43UnqaNuND1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Ac12FjNfzw2k6inxAKYUv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CvPKb9nJcSJRraUiHYPt3B"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2UFazbwZD7NPnebrdFWTg1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-33TtzHCoMQThJRdK4aCGyH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PjpwLqHEszK7qAMJDXvtJs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-47suseKgDUbpbMpmUs35nL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JR6QDXfJx17W8KMwMG87LW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GeB6Urm2wZZSXqf26Ldb9A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Luyb23GCFAKk9hk24cr48L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E4FxT3rsqxLrpa17e5ikQa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFavWSQKvL3dqFNpLENSRK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V8jytAmCYpKhaT8hVVkRX2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-26CpC7f1WkaxC29mC5hWQj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EYiRjHtXkribSe5YZZz521"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CaS3WqVVSjZvj8VvkwABwS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3zr2P1hgViULYqG1oB8bLY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QnLSN27qHitcNSahv4RxvF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5iY2haCf7Wd4ouSo5xbAPp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CaM9E286pqEDVJFmVeysrX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GJNQicahJcKF9QF97a3fcd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y2aAS7qi6us9xQbNaVoz1A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WgYZ6avzthwrEJgcnNE4A3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HdeygCyMpsnaFM5sEpJpCX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AoG5qP9XbCdQaSrqfqfTrd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DZx3EKo9V1LQYH8Qqw27Dv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fh1VHf6tasM8VRa3K4eZFK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gp5ibgK2k1sQctNNo61Qus"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2Xezthn95azccMZVARVKty"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KPmfYLvVnc6SZtybUUgJCM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9nJwWBegqSMf6BCKRjCeoQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J1R7NZ5FEnHJUArBZvqv24"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8JV4ygCgwxrys8AJFX4bfU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FJzYHtXXXfg8aQ66dniPEd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CMKMSiUCW6wLTtjtgpHGQ9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XTpXNR4NLjX4fH9QsWPV2C"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBK6UR9L2MKy22nvLf8URG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CzH3sBQmQNuCJSW5p8A3RN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ad6Tc5QcgPttHe7BHpnSpR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MJg48DdkxBtNSYqLbfF1U4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7MrReSPyUgh3Je3usKQN6Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-91oV8ktNA28JPrYnc3CD6t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QE1Eyi1yc7FppNM93gAGWs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ymzo5ibsCG16F49dgP5dp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L3T1GsAKc8nHrJVf9USZFv"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2YaYDm3pUnHnpvVffC3mLb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AE7mhfq4s51RkrEVRiMLRV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UEHvzDpzNQJr4uSAUCnxqx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Jgrb9RbyXcYoHttPvfL3Q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QmfKmLH3FSUSNintMpzRKz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EGb32VTpgkDiGLFcsGyTpM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4r4Z4ZF7fJUBGN3tKMY5CG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JXkjVzdQEQYJrjuAZrZfm1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C5PGErrmYCoD3AQTpQaU4U"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3vUBzmcgZCLEiyRvVd25NV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FdE93gCfyYDBTq8DySxhpP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-29GM6iLVsjwMD6ienXZofM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7KD8sA7ZqNVKr2NqP6rV6u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Grjn58eX2iFBfwJpV8tPTi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WGgt4nLJCQRhwzY7B7P8R2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ASsWujghsEZAEFeC6nDLi7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LbsyGF25RV9BM1GrQsqKah"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HiBG9qSKRHGQSkkJsfuVY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UyZf93S5mK8trvNhzFaEzf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JHeG1LezoLAyxbS7YyjEtS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HqymDFZ92n2Z892WQah1fd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9D8v9QAA8N7HhvDYLJhRrT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HbjmRjZ6TwFPWwaHGdayog"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Xa6f3mmRKapAK39v9ScaP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bh7B4FUbKiv3aFCgKs5HMh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KQJAoV94mLMmwZ4w9Qc34H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9k8h3uo95XgzBhfohYfH8a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W6eBL7RY1qtwVAC9jJVrKf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y5Aivaiq1DBsq7LsJw2E6z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ps7sBBsAmc9TjVM2WXiAFe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LHHSaUv5Ae1qA9KzdqteGb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KyCzzjFrmHnQmyHzoGQcNb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Suv8G6tFSko49Y6Q74kz3m"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2QDGjKy3kQrqRhHbz5QAqH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UU3evdTVJLg8GJ83rVB8XG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-888uaKE1FAzz5iCJKJz1aw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CpXKAzxVRUyD3nVxezMXy1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HRo8sRgaoVeyXgzuTUHRRq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GphMqWiDCwYkZ2Ft6TTu2g"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WfVtrG8uk4hxjV7aV5sxoB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73ikyB5VrC6WLgeBAGRbVy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C5TKjrdExtwoUXxKFF5Gdh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ARmyd9MKWNppQMDdRVNpXU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y1AoMwEYrhYU3pgV6ZFy2Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2tgK4phSUoPcVHmPcBkjon"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TpHTSZQi2w5ZDz5J4RKeFY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EHJKGWYYCBHLhQRpEJ8onk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A28q3LaXsVepqZSUg1zpie"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-THYrmZick8BV6eeBzVgnrF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EMoG1U3DTEE3LvgbBMgFNy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RebXXrYyzMfDfAhUuwLR4d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CVAcGuMRHdUnqCUMS1nC1x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AcDakGLSwpqV8iXufWeZTC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZk7trxsQXfVLuoSB6jMQ3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KuXbYn3XTEvmw5PZ1KGEbY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DNSzesWkQ8aayXuoHupFNU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SEQnpfpbbnKRfahCuTvRn6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CXCZBegE5qrQX8mrY1GqLd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-72jwMhtGb4ZMJnwF5qmtwH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QSKrtHzckhwZ4LiofQCckB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NSdoMSThDWQQNnhECzowWM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D4NBwreRfoEUdoSQ4E9PfU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LyGQajvFJPnTs2k9v3XC8U"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-REJTDZZf2fCRMPCtdJfuYc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FqD5nHm5R5C59DT3DgS93c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XU9D9jwBsT4nXG3CS9V1kp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1gUjkP4KZ2G97y91yshNhv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKqvJLr86mvGs5XR6tZKHq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-USYFewcGo6TGLeZk8KporL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-37wa5yrGh9jUhPcTfTwRBv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZHhHPucJwqgDYJxUHFjvE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2MaKve3upbEoRAETM7chvu"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CUWe22ekfsYf8pqWEEuiiF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SZDzVejHW1qsBAru8Eoap1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P1acUHk3mQc9f3TFXdc6tB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NjbuTA6eVdcpt2WkVioZZk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RBJYn7MinD1yCVjVsYc6KR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KCobxti7NTgoZkdJnKwfSg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wb55Wj8oxAdL1a7w4SmTGp"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-18wWVXZWvQZqcrdvezQ65E"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AC2bTr56CtoFqFbGNkptgw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQFqH2xK97JXCwi1JveiJk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DZx1G47ttA1xBKQ7Nuvzac"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4DCsW1nCAkQWNPbyu1wiZJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QX3ZcLkyRBQM8sd4rtivSK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFt6S97EwHWgga2qs8LXSb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GJhH8KkV913vu9CFqcjJAb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Va87u6DG8djTG7kNaLLR62"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EFFF5HAPLkni6k7B2gnQEN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BuQTTBY5Cd6FGVoEPNsxq9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XZXZhcSGLLZvNQD5KyMmYN"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9FHiixEARgrqf6kSs2caBz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4GpYFxnXggWyHe8aEux13D"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4yLsxxmnm1Eypp3hKtGZW3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NtqTgFCDrgGaz8ZRL7QpY5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qv8CQ4mSg6bxsy1DUcZYrd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N3FoW1hwRhvaj1rSNUz3wH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-65qYAJziczgACrxmvGkPWZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-76XsmpQpUbUPb1s16RDmCu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LxBLWxSKYe86r4YbZSMdBG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CuQcjakvAimzQy126F75N6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ByJdmtXhPU9DY5XSRkyfkU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ZRo4RKen6mystPfeHHyZp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R5gDJxkyK6VpKmV4qjEbRS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QC8JZ7kqZeSo1YTJv8XtW8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TpcsMUDXP6snjftkxk4bMT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TUkaoERwpSTSkGXrGfKaxo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V2f57Q6douEgUPTaxMkia4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3chxBZ51nxbuquJp6khN9p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SuThLYuqKSHqLSb6As6c42"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UY3qcEqG6S4p9BrkqxFiy3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C9r1kqeuK3K1ZBfJeWHN9U"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-69EN4vbVF111tYtJiqK4RT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TueugoWueqmJWydwt2eNj8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7BVhccfNYCtUXbg3az2n1n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AiosRexkXWDXzk2Uz36xcY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKV6XwS9N4Fx5EZh3cxrBh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YZWqvqSZfwEtJXk6hdf3Uc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JM72poJ484C1sSYemnSx8R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VrMnULcmweS4wFCHUN34pj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BSyDa7axe8iS85dgK1i4ks"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-19o3nff2fRPHmQy4JE1zV7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VtyHjo6TPfYbgTyABp28Nr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WQNcpKeHdYU84fKwyUxAJ4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ErfZBFwNEo8WGRU4T37XnJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FD16LqC2aEQN7wm5H51ydm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ixB6mqC46Z8883iHcz7yF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NTWHHrzG3tWhazmXmuobtG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4xaMz5GxS1GBzsfsC4GBYm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B5Ee8gL9uHvjoucNSXU2Hr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QVQTeWin3N8xRB2H46cbpo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ZP9Q9h95ucoCVar4rX678"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ZTkbLWoctp9TWuVJ2DBtr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A85qG9NQF5XPRZm8HGXzqi"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T3sWFXgCGEqPgGRmCWGqGK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ThTzLU81q3BVthNe7789uX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WwoHBtmWAJBHwMaPJgDreN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UZ4cB2HzqiYWwDNGtrmQC1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9yczngkTtTnquh4UEoBg2Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CsJp7Wove1rLTnPmU3BDfT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8oTMxEpQCTHJdfBFxRJYFU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBs8NzEmJ8rmhdGxVT5Pqt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4PGnpBS5Tgvg7fQ9DdSSL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XmJSCRt1yR4FZS9iTPc8tv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MfSKQCyn4JUtwSkAeZZJsL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q2hEkH6nm3oq7MAzZCZFc7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X8WRF8BLmNKsAAWrekzGUs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9iVw1emoi1Hv4iWVjQdR71"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1CdWZtwj3YuzVWZTeer67r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QfxpYGNa2WjtFHqQpwJ5kW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jv1gwvamiqvmtBSV9YFg5G"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VRQVvCce5X5pz2nN4a5UB7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GsryCfGXDvBUiw2Uo81XUN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J5VwMaAwZr833CXY2CoK2X"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GbboNHHTnCgucgDguVHAAr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TrNauL3v4sKztnqJjv162v"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7N1h53ZrrSLMSUWuj5yBgi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DButN7MLJ1TeWcuoMUPnnU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2WZdi8ZRKwyEWnsnBWUv8f"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LLzP1a74ojykMW51FKByfx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Aj37U5GRy6sMsHvjM6kheb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Go2kqeKYf48EsR9AjtP7pp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9uFADGai987gJogvLk6d8p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QNtfR66RSh7CZgYojkeKWZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GCY1YqMrNRV3eYMhTceA77"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DDx3qWEvs17a5QMUEhupdX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1MUkDWQwdAhhB8cdHRqiPN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P26rHwxq9Mj8S6HenXaByp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FQaf8FnXv7PiozqWrpiBpH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EQ1wckbegBfg6gxpz2o9bN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TTjnE2dLJwCfnfnJGSCsLq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FpSNx3okLW3nsC2jjpE4kz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LzxVe86u418gLvqB5iJWCc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JUcHxd2zSX8EdrwWFuk5Bj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VXUXfmXDnVP8NQYjieJHeL"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GupJJkSBfyoKEaVXRi2szg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1W4wfUbxG1b9LNPtMVfCni"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nqw5S8Z28Q9UkwrxMJYcyT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8kovLUrdL29SdaN8yRbkS3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N4rta7vNuG7zRpvL7gyvKN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KWX6pQfn7sWu21g7UUis3P"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RT6wceJqFjgztoBHq5HVYk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VKKqkMJ7LbaaD8qJCdFLeG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PiM3CdDeukrWKykSdtfyvT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9FCiE8T869Vfzxub5LMLhZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7o3iuTM6RwE2xN8opoYckb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1GRKXSyoE4kY2SGjEiyVCj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HbGceYmkTDA27xgW87vQEi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J9U2GMaPtrwb7Q8fPazKvC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HMwo9PP8EMJurmQMEtsUyd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6rZ1Cms1DMcQapbyhJTdVs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5MLJJyqT3SumVF6MzgX4wS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7r5wKVRhsYKTSWwvZvYkmP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NnRVDhwY4yPwzVzv6JK8TZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8oxNAVeHBvKYWpu48rCh6S"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VfTftWS1mCVHA1NstyXuPH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1e8Z8V7ZXZMaJXC7x3VcFa"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6oVjULHgA1hB5N4BMzQc7R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XFJkjCM8XP47WFXzKCScVy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q38xBW17dQKHaP8QvzqrkU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J7sFPzb8EHLAvLXLo1PdqX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gz2Hij2Z58vJm6FZxynjB1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MDuZ1hQW6qMTfUNNveMSxr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AK3Wp8AuX4krV8EmuEAKBe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RhcoKwsEf2E8beNYQHxqGS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ey89LkK1FJ6NmCcrZQFm7Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UzYUNUh3e2fqvqKG1KSACf"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8YQpVEid34kwwQyvPVSC72"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tp5WYWKTFy2Tkie3YUpZDT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EsAzk3J4eD9NYnmt3gE7Td"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FgfnNGLDdHEJ1cjQq3RtUm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PGQWZKhTrQd9DLPZvdu9rF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QKYkK81HKAPVuEFZNcsJd4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-65YERJEqqik1eLqXhceckK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LwUY9R19gXbUTzRDVzzdyw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JEqGL4gXyAC2wDxicBBDnF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9mdwjsGpF1bBkxCBnWJjJh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WRVME7kYsGgtWZA7AiqiRw"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MLUCjhYLnmQzaFyhgmBX7X"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-58V4uWHqM6rdCbEMhr9PiZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BReVPukTG9iTMBYQdAsznB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TBvZDxQ8N6UvKuSwmtcXyH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4pkBMV1CKDXxXnR1sj7kr3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Szomx1rwkHvGGn7arWywZh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-64s1qC8aZ4gYsMEJugpZra"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6FjcVdfAwvu4T2pqCn3zSP"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ki8aiMsh9s4rQnHysGhZk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H8zdHA6PvdkXjHfUtu5wdz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cyog4C28QkM9uYdP1MDHP6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ap9Y44Kuybz7qj9XkXks9q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D4UB6h53qp2qUzNWJPgudh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NBUuvnWRRHcbY7iVQ6oAvZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EodJkLbpSfk1jqJWeNJ9qH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CZaY4KEZWXDpiVZhXS5SLP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGsGKox8qeeVzJ67xPDs9i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A3BCNjyD1yVMyvcnniWnGX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BzLABw622V6i486Uir2yWA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JJrh9p7cdBX5TWNQ6mrZWn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SVUJ9mKbgZtJSMLzGboqVD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UB2KehFs3CizgPCtuZRDB7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fi2KirW2naHnUJg7jorNwS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5bXZVuwNpKJoQDPQGVq1eK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NzFwXvLWBDepiuBRfpG5Gf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XkCvpVm3bsVLFsjVSJKjPV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AGMmVQ3DsSSGN4ny7MtnMq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XXwDtqcqeXpTzMbD1QU29A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UHkQbCjz9hwf6gfLTVZS2C"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TiYSsirBbcnjThnG3yL6yA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EC6jcPkCwCqHdGrkjM6M3J"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M2U5gpgSnS8vpJx3v4dxHa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6qd8RfqwxJTT5SJFVy1rRN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gj6KCEoTwPCGv4LhNsJUyg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKPdPbnR1PnnXqUeqUNqfK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1SK6qrgPkGF6LsiByrdtpw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CYxEWLXTa6Cb7CDpUwQ2xD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBjDfcyPsMzWLSB8Lovsmn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6bYrTccx8m8jMKJLD1oBwQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ky867LJsgKe7qPGdX3cgfT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Us44pMB7ZxCYiCGU2YPG7"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FRMbxbaqzByRZnAJK2G7KE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QC3uSTmm9hozLVnBmxnrup"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LpKAdEofhNQHHZivoEZMre"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R3GrH3BmXdjsp5bVVmBwa6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CkZsPm9PPqBAgAujRrLJeg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WztBDBWKK34Z5v5kz9M2QV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F39H9YNjAudGCdf9kXFRYr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R5EhAPo5MEnTRU5Nyt1v52"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MiwWVQUVohTLDM8PPd6VNj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WDoJgAzMUvebZtG4r2t4Bd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CrMmxScXGcgWBZVBeTkN3P"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TYTTibvKuy68Hgi9N758HC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K4Ur64tYnd3ASTvMq7DwBB"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GawqV67yvagV22J7s3iNKe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Yb2zcdVC9PP1Uz6AFBFnHh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vw9vUegXsSDdVCsRrXFyU9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LqQsMzxAJhyACjdHoCxWrP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SubB2VoyZTGGNyNCg87rek"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NdykCPGA5ofxsok9Aa1bms"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CVsMZ224ESPeuBCuB6q7dU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CUv4gao1jAbp6AdpoYyqB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-98K6yPbU87w32yPdTjjs3A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MC1RFrvivzEErx6iRfNyPc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5sMh5SXT9CMNyE3KGtftoP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GsCb4dx4ZZELa2q1c2UvJN"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MU79aPr26AJGt58ENANVE7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W6ZhWYBX2eyfa1iWzkncWz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3vncvSVbE34CxuoML2hnJS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7AmFJ6r5hA14BZiRyVPf7V"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pe8a1DSgFFMFvybv6Va6AU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DgtEA84EHwZ5qhxfeSP2n5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7szBhKGUHfqihGSvvT9YDg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DB1QmsedxERrud7YZw7Egy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pemsp8EJtrJyBFiVS8xcUr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FRDYxbQStaw28ejwACzzqw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4CpyukbQqKPX3FHWC4zw2Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SeXyBB3CAYceuJ4i7tseFe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BHrezqjMZmszzw6bRfBNYr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZxdFqs4yDxApSYtnxK61K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VEnMZH3mkrRKNTRGwtnh22"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WomeHkbuU4i1XP4fEe3Xk4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y2wegictWSG15pXcNsx2y5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WaFmNj2PeecEw5DRpf3TcX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4r1WN9DTTCJg8eqvp8u7Nz"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2RfMcDfyc8WkutKgjeTYsU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AVfoBZtGgYeqS783kZUZRe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RbZipHUMoTsDSU9xCPxdLX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X475Zn7U4niwD1xKvUVLGN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HuJbC8vntwuFjWDcSXRHzi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AnDW5eGprFTtcX1M1FTd5Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1tAc1e4oevPkxFjAZBzzsv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SPPex61mQrPq19N4VfHqtP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KVqpLzSxjYtjZd7CMY5NM9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GgB3koxBD2ggtALkSujtJi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7knc5UC8mDoVCuuf5hZkZo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VZZkiXdn4hiQdc3B7iwDQB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-23RrHxvFpSbwDoYqqA64xD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4JUfiXFRGbopSHW6PBRDfc"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B21fzf96YAXXuunST6mzj6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-34bHBrwzUut5mTwY9ZS9xU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eng7vjUw7dkcScut3YMB3L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D6pJyNQPJBrtnjLEtLhYU5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KGqkxL9n53VLD85txtSfvC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VhmpKsofXDGCbkuSa9LkLw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YQL4BpzSYmpo4QdWFtAMF1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-45aqnvPwmvzBBAGdSqayf2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sx6YAN2hUop4KKhNALW62P"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2V12CpVmVSgHVyk6SSxN5Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q9eWH3jScDGgXvWA9VnDvq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8wXW7bAVoPLFpZHVJDhycn"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2omxuTbnx58yu9kKU6Ebgc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8pNcCuSfRkR8GatnXHkhHV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TVuzkKbkJQZs8BaRs4Srss"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hp6yR1iMEcz22V3SWYAT77"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NYCCbWG2gGcX9Wm58LyXd1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2xEussDeATL2pA3HBSQduh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4HQZsJevexbGyZig3VjS4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EKpQDRTev2niPmJZGZ2zc8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W8Jn2XhtMM4jguzk1FqUdX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GnCFxkUVUmnCSJzTtn9chQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8wRreKFHwy3HxN3qfSEjrK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Re7knAxCm6CNcaanHeUdqh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U9yzKQ6m8ybNxZvz3kR2TF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SPzwB8kCeLSoCXw7G2fJGh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EM6DT68vP4vKNn5GAXAP8z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BWRxhFGHSzdzeEaqtUDPAb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CpQ9AH91gzWPebF67uNLQx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KJLSQiTRg5ZtXZ34V2WBws"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5WZ44oYrfmdXPUTtoMPYuQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ycSEdqKdXFqJB4bG9kXLq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8KGAp3s4NobfWso595tYZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8kbxKfshshoHw3BN2z9k9T"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CdavHcbELrcSEynM889Bzp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TcFGhx6Z8Xvhx5SM1P9i4H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UBqb6wGiCp7M93qPsSFN1M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MvcSgahwxQGzKxJeeDy5hq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jvhk1sLqGdFAUVrPNYD8df"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5mok85xBbVoKnr7wbaGz6N"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mb2SSL6QnZxbsG32JQ2RrS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8oDR9gwjQqxtUQqzTBWYdL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TzJBnoSNFvzrDMf2PE1Y4L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7JSqVfxgpQKQpzJcG5skhU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GH7PfMqGKRVXBaT4doVgCL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QjfSePLXcg8P5B4W5ZPCbA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F13aFGTm1UQoowjoUdWw8G"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HcrbkZnek5BGC437sxBWmL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9AiHAWLAUSuYsAa4QypWaM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VVxTBF1HVGEd7fG9NJj2Q8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBjUnvdZv9q5zeb9Umdajm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GpAP77yh2dBiKAgKmWWZy4"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ps8PdigxcPzfxcLZ5r9SHK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JGkH25Dep9LFgrwsc11ehW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EAMsHyuzqXs7QuHNnpWigk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gsp3rR32WGGUfYFkG55kf1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9YKNSqzwzYnLiPS9wZQzgG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N7tT8mL16nFqrqLcjEWdKo"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1GeRuXwuD5kkchcs537nx9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3t4yedtgJ3GEF262TrrDtB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LbrvcVSMfFBx67oev3xpYh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MWozwzJafaRK59xv23eBcX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N7udc9LeK2uJVLpynXtx8w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKQXV4swvFK18EYVJvrtXn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7pcgWjGdpxVoq4xJoqMXAJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RFDsTfr8HBHDQx6kpXjnHN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JYazEujETFNLabpDGQUijW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H5C5FpeXzgm9eAGfYuiiJe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JE7yuEfvYT5L3bP7vMSbX4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XJjZZW7XJhq5ooY8hkYu6Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKkjhkaMDKNcMKWBjA9Xz2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F6gkZQFT9r1XaEZVqUk1bH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nv4ehXQYVuNBXbvD1eAF8F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1GKFgzUVDLgmvFo7v13Rib"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YMsSfue9g7CSEXTEbdBdGu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SxUk9sXxChKKCVs77xU69f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5zCtZetB5HWgtcWYZYKDZj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CmBKgvYXTwgXhEVvaZP8cJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9FSh1Xwc42RzcN1goaG8sV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GRhGmo35vd5d1iDaTbAKbc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-66fNFigCAumkeQevrQVVdw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2t9DgwDzXKCjfDtiE1tr6V"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-56FqDRFLYeP6NTbZ5HLVEh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R69oz19QrL6LrYFXtPWEKA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H23NoPMNhQj2nDTKZp914n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ZMqksGQoEAA85bTkPSX8K"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-22fSPvLoXuJmYBwDRv7h6N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2CqXtYmArdT1orn3ey27Rz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9deyaAeD3LbGL4eXqW6UGp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L9LZJj1TUWa2WREEsk5WWk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ADGnvirRHtwDVTnqsBtWAF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDw2E9xqmQZn2nu9N7TcSm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R1MwasmBN4MrzdYMRdZtwv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Emd7efXJ6g3H9tnFfeVfBp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4P4JKAvp3zWjNv35YsTFzL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TjTH6QUU2B8P3k76kM1CyS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5h1xc4QKHwDKxWWPAPLR9G"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RRPa8Kpej4qK6o9viafQkt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8v7rcoBJfGdbuLfD1NQxNs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RDUQNNFfUACwzhTTyjj5b8"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-872cW4jR2sPzbWeDejP32v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3CUDTLZBN4gd6ZZd3goJRB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QZgvEuKsAazyaCHLALmgJB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2zbhFWQhFdL5bZoet2Jmo8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FUfpniMKyD2qChL9czYHRg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WcP7GrSUsuL2UfzakhCG6z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NHphsQ2RA1NzsA1nbM5Evm"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SyM9VY2TovHxHbsgk7KDJs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DQq7FH8QW16BAtQDcPj4ZU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RFTng79RzJbnZ5DdYGjLdW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5DamJXk6BgcUj4ovBVUG25"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V1fpDVsKda7PRxEsEVJetp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ByvG334wqvR9FyBwMkQxi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WWVRyfnvyC4ZEj1886ktqf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-68oV9XmvWrtK8sFRfuT742"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRPKtAqDqrLf69KUpTgaKk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RQNQbsLnSV5VyAiS7QwUkj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5FHTdE8jPVVxjgTQCgFbSd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eyz5RgXr84XhdYfGQFsXhj"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Dh3WhC4N1LQpSuKhr35Q3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3JBcwQs7uwLxcQbrKwmsbU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WjMLVY4rLW7VZGWQwM9uqD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vpt48QdzHcEUk2uo6ZNAHy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YCzY4WGasJBUN3XxGy5c5f"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8EdvprjYTnXprub9MbVF2Y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ptZvYzAruQ1hq8xyMefe3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PRGhBfjv5xqiPEdn7ft5rD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JLaPfLTSPJAbTqegMfbqqo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QxEqGKcDmRWQN1qQyenFTY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5BzC2vZhGQTFTGcPVDCqnG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T8xjU3ycTnChz1SaR8TwTg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S5atQFganeHnnR67dDaZzS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HZDUy3RCsUtP9ocKcK5LMA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2RboXz9zCenALoiz14W566"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UuoR5BeEF6tSrpDSc6sGpB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3h6ihnUkxoXqeSPu8ZYqUv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T3uTbKcm4dsnrGK2GKkyrC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KYwJS9U43eMuieXi4ahX2w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3CtALySKr3Cs6AAqYRJzMG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D2K5ahAZxaayBn8e2JFYhx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MjFt1ArRKH5ZhyMBDbvB2s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6bZgrGokpVsgAe2JVJcgq9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3CDMAUrEG6XZLjrY6zDsrR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2UWEJGdoewymnXDqMCcPGK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-71Lx4EAtt9U6oRpCe7g8cS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3pWjr3hf5c3XEfGCSRB3hc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jm55vBd5wViuK8YFWpt1To"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KydHKkCtkYHzihzLLZLoXs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AXcpyy7dy8jPd7HwV9Czz7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KUB6oY8wQVnGHZaWUJwAKb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DYCRDM8WN1YeGKsjA3QyHn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VpZYYYi1pgCRDKtZiXV65H"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8j5ydCe5ZXep59yJLiwAKy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NvdPJ474oT2hPdHB8u43cA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FZZdxpVQgFaog9AmaAN8tT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PGg6kPJ5Ff2TedynoitFyK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KjdhYKg18kzGqT1jbga2mY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ki4USb14zsn7uguqrybkJ4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XUjdSBSMgone1MZBhqvqjp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AhHcZqZCkGfFQLzuizUSzK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ZcED31jFLXASPR46bKR71"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W1VetL6EoN1HnVnMFZkQpN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GT4TdWkMeX5eWyaXVJon61"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QYWnujKT8Yntg3RpGDsqPr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AEwedZ2qR34ho5crhJoZum"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MerFSpbgGJf8e3Q9Lpc5WZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6JMg8qbtMekddScbJ3Z7WM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-97b8HJbruV4Dyf9r4vgfRC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UtPpyRbs6sU9vJdhRorTha"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9vh3RQ17syKkJYLMSU5GyM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6rjfnCb3AGRYAQ1B1noST1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-THubRtRaLA4irTBkKfEVjB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ZqyHMGuoFcLQKpRkf9A4p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8zHjYSXaV1sbVohEqJ38qJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W5qCpEBmQPE91KrobcvfyP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DkqH5d7DG3ZuMThftYsETM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xpaehw72rEjxqA3EXEL7SW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HWhFEgaQcHcU6L2xG1ti8z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ht8wXsMUCiXMdie1eYka2R"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5DXcy2CNzagYGcZftwivzr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-13A383fXmyrSVCP1QvditE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8nN2kuhS195vrzxihdQieu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YLxtwPR4LcXLDz9y128qaH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JgZab71ALZDQL2gBFWkFxn"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RnazRvn7tbUPqkVyrTLEG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9u1qz3Qr1RWWAhsgEZ8Ffz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ScvrRfYePEcj35c1cPMMax"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PqnzDXU8JkGEdHNHuitF93"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FbuttPSBmwi5sAR3joq3yQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NXqxaET9Pw7APGB1sJmKkS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GzeJP7yW1JTbiQ1WZf3esZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QzcwZ9VhwfdjZVWfynXwXN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-75kuAFo34cs6TsDQaz8VtF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6yWX2HAp6rhUz3ryREERKL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9XLfxBkfxgYZcbsMT6JqYq"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NUAKoG5CbeB6mQudNJ1o8b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QebxpiikCp2aqw6rgEEMbt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2AiPNCPNXv4bVfX8LyaFyW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U8sSxpjC9w2RqAVzzCt3Sm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-78yDtXstbhVNt91n34X3sS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BAQrrjD7mhdzLSyxqgXGJc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QzeZLQqkDHQfe6nWvY5dFF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4cyiyqj2pSWMPLV1TD6CVd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BQAVpetBAxkyZFtoMM24i5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-URL1KcLNdMJ4gwBoYVLG5A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ZQqvrNJ52nUkUwU9n74j2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P1fR4sj7fo6rfhZhR7MLvr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SAoJzdE6THQjcQAU2mySTc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2rMrbHKuauFhmeHH3g1nMC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CuGGg5ca8M2Yx7xnj4rySH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Js7HvZf5fyvqGJvtu1UDaY"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GWmSqvewLbHJgPkBcWfCpA"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8stWXcjgUS8YmZTWo7j4ZN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2t4odh7cLS4s7Xa4NaMdku"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9gkz5NNCSoQuhsFJG15nFt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3k5wvDaRF9sDCtDpJ3nHB2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2HGewyXfdTKKAAzBbPyNHJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pykrb2dgfh3avxHMAy8XCR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1VjNiBYRFdj5aHe7YETRqq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6gJibLg4rCdfUZbsi5evEj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YPK8RziYX576qJFRxA89tM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CYLdqur3KMsN52QMZyoCDk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VWGm7Q1vzjFvi2ubNvTP6u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CZnDhpEvmhMF4gMSDcdjBn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKgfbjCchVg4J6iy4ecgB9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9iCEKWYmcH7HqxcF5VzS5e"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5TkUnJbjcNGcnMLL37QJrf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5pri2RhhFyMrSZD6o2sypM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QHaioDcuk8cp8S22BoAFSn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8TiHtWDNJ3HoV5wJYeY6rT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NK1SU9cJPS6mLxwVb2dgXM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U6bKgoWKY7u8fQWsniSP13"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HhrUKYky97YKqugiitzyhR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mviwuex6ZSc5LUfEu3mpgA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YGGWZJQB7noJdnKAMemP9p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CvcvTFvNmUwq78uLRwBVYf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2YsHWE731JyzvAaFjrC5aD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fnfd89qDMwafQLaD5vFSw5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XCGBqhNTJLZE5eoa6G2S9C"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C59DCy2CfQFQudaDCAzTij"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4De36nTdTfeC5XiMoNvPUJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TJFHr96H1y2TEiQ6CHrFqW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MtiaWVqwCBTyv2v1ZK6Nmw"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FQeL8PbyhQ1uReT8pxfKWr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKDRfWCeyXzTKHjWRjJxWB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UfofV527qvbiMqDjzyUDMd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R9knqxRqJ6YkGMmsxTZCL7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RhSnxYcMRG7CRgoWb4ap6A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KdiK3m7LtYjoQfJQkY1pea"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ivzUU9GYH7w4XdFEYh5DL"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9MqUVQMS9L86tNyAqxtE8d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NxstqLKfrWCnau9Fv1fu7L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hc7w63GC2LiR5btZSn78ed"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQFqvg3E3VYXb9HAat6SoH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PEi1otTAFabFsQSRNodV3g"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4cuopve97fVFhUNZeeSQmw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Pg89EZ6ykwRpU5pGJNwTU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQSKphJjwWHSTFMYjuaWpD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2JZFpPpQ7DgbhVcMZ2ugYU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5XzT2n79BzuAq4JbQs3tB9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ekt6PoFbnwFKFuqiSm8rbb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FwEtbFND8w1winPaVoFgdP"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5uMSAftgFaZqCGEE1cc2YC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JLdYCBL8PzYQZgzswoaEx7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U2TYZ223qxeGMBiZS3xpCe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2MGDqdxCxUwhyNy4uyyDvv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KLrURG4GB2LAeCxep2SEZu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5zEdJi4RguqMSAdVCq6ybn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LRacdKtxgBBGUZAdtmR2y2"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4idt3ZqVjoSjh3pqip7fcf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ZhbezpB83rNew5QuEQ5HJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Vn1zZNwLcXHMffGB2eA9o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-13dfH7eYUtzrhjK669o4tf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9GwqqtWATEZCLmPAiX1NU2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VRNafBU9eumAhi3kHAeVd8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JUS5Q3wjYCjQFQXBjnvYUJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q3btzSw8LG9Rq3sNx3iGH5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EKptw9C5qjhzcNB6egHmgJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FjQUHVmuANjie5HSSST4FE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BHf7Bgv7TZxaaAvvzjXmjF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BQVm7P2s5Xzc9bMX2QF3Qi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BWPCwpTtPwDoVeVbHyUtre"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HEGzGCj79Eq92dbrp8iLJa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JkXbA2GTQhvLqPdSZxRYAJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J6Zf8pfV1N1U34CinsjLnM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G3ETXQW8QWGkEQUdHroaPc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2KBc4ZUTTfbQJ9DSuCSfer"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ssip5WAmxkbYdLf1gnJ4jq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UVAF7Wc9zY7Cufrm5ueSqQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KnZxzQRjH6eCStzwUvUpN8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Spw4wkcVtrhZy8EG59UKZN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QEbaA1HaAJj982h3xaFcLv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S2ZEAjM4Xjsb23hLkStLG8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NpPZFjEokXG5d1bv7Z23wg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FeqAgSrs2KWqCUgsHQTm4T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1FnKDmK6r3HMYxdv2tMLtj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H7nzL43zB6Gtb79XCeZDbF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C8RaKBHEDczmF13iRQ84hq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8AFX2mLsyFAwDaiTC4FNcU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nd1G89twNp8uUXb1ptXhSS"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KEauKgRQsUdJFF6WxqUQKA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3rwr3oJg2MZzxWf4veTivy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P17XDZRsNBf5Akc9Pg8J1n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J3MkTB6nWpM9vh2NhJKe1w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kwq4EJiv3ExfH7oBxSxqVf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Yff7KFcfDb5uzFG4aeK3q"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RLyECgG8GntMiXykzeYpEm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BaN8HJWt2xkQvLWj4qCnHT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K8J6ahmpTA7d7MKu2aiTut"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JhJy7MmNpmaVPeGTHRq5Uc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JPeiBBL2NqovMDx369WeQ4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CyUkraR1sbs2PGTXxVMZ3j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFuLVYku6nYHwxdHfcJDVh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BkmZzy6SWWWjdQN7DrowYR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AQGdNxL8ujqfDhAnEVHTZ6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-34BvrHSja3nosKe6jh1mgB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Z37RSGhsC5Li7WpVumVuM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9t3wfPQk6p3rJna1RXzKgq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EE8XUQyh4jtyA4WXAepSvf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RQckDFJzsr7ox7dqHTNycQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Jrd1JitCHt3wwHqcUzHBS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BbLeQS95qJdZyWUknpE5gq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y9NthvuwWsePoMNucuAaVZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HBfthKDBB6bN1pzvkuVUAV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MeyaSHps1zA4GkqMToydRU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M7GireEM6aHzv2PYLPVBYV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hw9ugC49sHMykvsHw4tdWq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R9f8QcXBYnv1yzQTzGcWwY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RSJ4fWudq8EnRNySFZbWje"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9VDvuj3v3mGVkRKz9ejhrb"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WmKjLGAiz4qagR2DFETrKD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hq3C9q9H4HCY6zgZHfygjT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jy4DAiw874jzUA15CJQg99"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EfgezqfPpTfT6J3fb1kNZ1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RnkB75XQhLaAd1eQedtpHo"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YUnyuwh65XSyNrwW5E4Shf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H4BocGKENKr2emTeNMHZoy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LiMnijvsfusUhXVBjQsQ8U"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1UEb7BehrXCn2w94pYBPYr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PyacuWoDH4S9xFrPFaSf1s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RkykYneHCriLH775rBmQ3G"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VfZvsdQq2hEPzTuyrZxnc6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LAvKLzM33XJWn4EnK2mkB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nvb4aXEWdQkTJ2suXJz1iP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Ak8R1rzR6WQDeScxNQVyX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BdjoAz7Qm1yqcFB9wtYBZ8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WguRKsJ9RgMTfxiLnwWvaV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XdW7KzwBRWsZif9Unwro8v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XvsJPbHXxyhZKxaMVw1RiM"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VJAKdofegrHAQVPGZSCXkW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HEspXhRpYpL7MGdVokQoMC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T1dzwQJT2RUTrq4Vb2pxsy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CMT8fYDHGJwaouaPck1Qsc"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VgWdNx28iky5QSMmtyTkf1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APk1qCVg59DbAbzCodWXLP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-111C588ur58UnmNbKq5omt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QqiVjmSXUGRuHg2zBeMgGU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q8yaggg3rfmQw9F76SiH2T"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32B1nTGncndhBkivP82g8M"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FsnFDK7SnANfqqgR4ZWKiF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LpMKBJruUJcyLTfo5kj38o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1q81gVtw36SgyDXZ48UiYq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7qRdmPZwu4Q6CmHqkevSvc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5KbPnxHfsVNRmpPsWFwTyP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UAwhxzG8xjcaaW1nFYRcuc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CNKwJedcPnddzVbXGJDeEx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3C78yX8HZhZbn4PNKQ2zGT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ccerLWJ6HMiLQnjGopEbx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6nm2bTQXmhSE3RJKP9ywTp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3VDcpvWxpKaY9CSiEq7VXm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rvcf3oPzNCcWjVuikwT5m3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBZRnSSEcsRV2pkbcmGcwx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Df2b7gbvPUqsHB1Upd4ayb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EQxLpsvCCtAhbYeCuPd4Ag"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WQscueEqjHK74yhyVwH7sR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qt1mdUnSCLfqsjGQPGRKLE"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3h7jZBywD8kq2ezZma7XBK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T3DK1sKETy1JLHk6G7hCGr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PQoBKqhtCnYwpADPUBMM2H"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7rwELRCfeeuSdkZxynVwSx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SDX37bnHBCUWPBTZXSP3Cn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MNyn1Evc8wi6XXWqBYGzNF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Td4ig4XvwKiutAKAnAGBb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VndM17oppt1Q5nyAqAhyqq"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FDJJymARKSVL5t3uwa9mi5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K25TK1m9JMzVpc67vUoq1W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5bSqnWRifRPKMz7gpbVUDZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6HrcYVFnDkcKxJX11NcbYw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BhDVEb3QrtWgEgTDERpjgv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6QME7DkkX2TDKXdNtT6jUs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bo97EHKcoGZ2L7FYTHShFK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ENus5KyPyGBvA3uW2qmH1P"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R3jGTDWrk27Sx4fprYZdUN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VYTxyaFDYSvLTGxoVzLUkg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AaXVXMDuLXLYqtqpoLb2pH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8fWSCWc99jQN8o4RfyaYnT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LE5MNHxLCCtBT3ZuPAnXeh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BQrKzTQ385MGcRfZuqPYii"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EBLYUcgnWg1tu9DtVF5pm8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PChdkn7VSc3KA88e7QPe8V"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YUkaRpbspBnJ2FzvYMvfL6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AR1Kbrd1ALUt9zkJfupuBM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UtFqcKwNjwwpAmdYMEWJd1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F3iWjU14qgnWJ7iJkKRJDh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QKws16ND63P1MfAYYXDs4S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y9RcRrdFYbp6pncnLCD7WN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PsHQABexnPMCVb9rpCTqwb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BDnV6CwkPhyadkPkhZVcF6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BvCuCkCjFuGW23SuVYDEDd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T4xEysHxYfggSQA9pETKkt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RiDLk4Ftjf3eA9d1g4Gq8a"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SxHdQhFzXrTtphFUeGCEcd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A9c7r47nuefXQZZ4Zsf8Vr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4pbZZdBGumvHEUDWnPVdm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SPQ97UvuWW7vwx6YosBfU9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XdA9shocs4cMpp9AYtCuNn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H5xEXwrGTp6BLeD6miEPAZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-74X2GbiJzP5iChmDsyuEWt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9t2jvZd71JxDKoQhe6Sj3a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PZfgmedYSzVJGUhwj4kP4L"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ssg4eSUzxJQoakKBuJUdQC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7NNi9pkUv8uZPQBFe98KVh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4tHk1p6pqQBs4HzA7R15yt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V7yH4KtLVjTMzg2aNYWmur"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CVo9CiHyw4CYHQkHuTokj4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-36rHkHKS3mz6XJadjAZRUf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKxBTFJTKfZiEzyqBDY11K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6PwDxJsJEA1to6SMARKsuG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DzWvF7FTaMKkSkxaVPcQJ8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3PidPpfBQKAhstQZLWRLCV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TMg7pGvrE3swUSbWAxUZ3o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1xPfqzG39yVpxbpbT7HUtj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KYHdqaiQaje7ooqq3yCCfC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EU3amJ2coryLmM2gEriXht"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7XLPhL4ygR7rNywqRhgom9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XeXrD3dH4uKXutyfT1bgzo"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JVnJYygME4GciULquaxcsM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KonxhiwSMR6nh8hBYivurS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6vSUD66CpxrcYtYbTJboQA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1FawdEkYij7fuYKd488dZq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FgC7D8xyc3rzb5Pv83mQc2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nd5sinL1j2PYkoQDWTBoDr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2QwKGoYwogLgGcx7yrbRmz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2FzM3UAc9ELSpuKeZAwZtj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nepe714cceWtc7sv8gBrpF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L7q82gZQQYsQSKhEPBBkdB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JYRh1P5KGWMeMr9uFpDBXg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tk2R41WLZ2XPxuHihnw42v"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PCXMCSMv4jMPsWdvscWAey"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1GauhcdSRQw5TLQv6SaKSb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QTivdW5yeEkbXfVWwjYoob"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WoX8EVp2GPCf1K3xZLXNXy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJMuM1gsw8UMuCSwyYGdRt"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SwdyeQUTzf3Lrzw1FKkR8i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8a9y68bVaRcTExZ7o3G2XX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RkZw13PFXTruSXWLpkfUJW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D3m3MP7gnXomrHRybBgrWn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VcHguSGACWbEUcurP7DVsd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QmiEXvVS63squR3EFhm4iN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Nt5oMq8rmZCu7s6yjD3BA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6wwwD9sgs81CvNBsM1JYfP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nikre7vLZMSPq7QJryZMP3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-44Uf8rByHjGWHE6Xsimdb1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5WRHgdzBdDPF3itraEiiz6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HCQyNtyHif9f9o7UcVSoM9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UCMGQ8V7tjjGZJL5dmtpnX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5wZ1ZM2woju8TV59dhjbMS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8bkJaz2aPYfKrQBBtsPywc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SsaKCMdx2yTMHRTe3GWCgT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RqvPg77WwsqECWcobVbJ6G"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2AeunNzibgzgbSpPSkzCLo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DtGGuxhh8t4kK9PRRdfAnR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HwKWPiEEKCjVECMqg42JbX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4aE8tT8QQawCRHGzZMbiwa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XeUd8krFXMM3aDDfJdpv6t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WPUeMn47VSmKr4MsCMNP6T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KZvqJZCHfLRMheS1HLoT8q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EvCG1uXw2gjXUggWx5XTGy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LSrUiM5VYvqBfz56TuQqm3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L4mXU6WgwVgi9vQdoswKpU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A818bkvBbHTmQrpsUxPfkM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HS1Xe5SLxFkeHHWJM8hLzw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9e4tPDMDQU3C7DTYfFck2o"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NYLMMdtX1CnJvyFDDrFU8V"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8gQLcicivaz8QiTMSDYrHY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FJYJuW3vuU45VXJmRjJUmM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XZwbmjdzj9dRDruaevSFBA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8zynXmPEui63q7SSkxr5yV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P5VJqUsGyjGKaUTHrpTvyJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FuuRrTHf5WNRHFNu2itr7N"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5g6QbJp3YT5ANiFUBa1Eh3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RhKxc3j75erHBswjmfzG58"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5twtVMdmqeUQcrEaWHisDc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ENGomjWkGEBjPqfqkJiXda"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VSMUfhASQ9Zq4Xc9z3gMvL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NPUzmGsYxHAiYz2JADa46o"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2tKcxvgjjDLrjuCZwAuGpU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BpiewPuQ8b2rov9HFGzvkh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J6V2noyjVLVRu4qJSDVeUX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y9rfavRQ5iLEdt1PgGceUh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6nL8iyAY5BC3UDk6dKXx48"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YYiGvmpr1UhvNsgc7ftWSZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VYZKm4uVcJ43qY27QStwQF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MmcyShaY2m1RyFAxV4zkP1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E6xLJjGhWfmQLazpFv9kHU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6FDEeD7mWvDLitX6sPzB7F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5YLH72jZf7yBw6UGcYaw4j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DCrjfzyz9prSMmHqaTzPzj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SJ1fJas9UW1jyZELUBG8C3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6pfuQrExKMBsLxABgtov8E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MNL8WrvrcYKtVzV3sYvkgi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G5JBFSUMnLrwTFh84mCVeb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Pn7ymMDp9AfrUuFqXtN49"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-559B2hFWQLzc2p8AKHGWyr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-USP9xS3jzJDMy5LBE8eYy5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D4QKazegJFz17JUA1x11MC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6BSZCW2YZr1YsEzrujKqbY"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6dVdJhq44TnqBB6Xx1Ej7u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Uoz5r4q97t57ejjaDGNMf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J1TYLMgNKXRZuHxBo6Hq6n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KtE1fvwjdQpFr2SxHrs3NE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BnPnmBEYrfkRpow4hb3NJ8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2WD5NGH5SULTSYh9yqBgAP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VT6zYdUN4P4197xqTs3hRh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HzVEiYVH3tB6YpwRcwq2kD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LKGUmF3KSSRan6pxLtsTLx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XdkZZXaCRGtqvvgVJML8Fk"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6H4oGpUCfyGHcxPWSQg8Vp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NB9Yr2bAy1gkjQuKeHTMPo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DdJs5Jduf26nXLDgv922ua"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Noht1T8MFeHih2yG1p9rnW"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8vo2QLFkXQRL1BqhQa8Kx2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5eTkCrfH5aX1ucZ4P7VFsC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XCHox6RUZi7ze3ENaW5yRk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R3Bh8maH73n8cRL1RoaFBM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EiGCnL4Q6gaNWYCAW7oT9W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4V438Q9AZ7YH4WZqfCKhpe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALbLd7cfH7LuSYXefphigL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HFvvebL4sKQdB9aYZQ9pdA"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RUFzmM2U5uxEKr4cDhKLGR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CB7UFgLnBf4P7ypAx7bhqZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gng971qj9yzptd6BiC3ja8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1JUesgRENw3efNPAsLoNDw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GnNNMyPpTSr44QK54sQNxo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EdGqRyEhsr4GRzetcwfBwZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M2nESsHKSHayjvedsZgqpJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pk4N6MZ4BcL6SQWaFR8n86"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Aq2mR6DRaEcZoWGmPrJpP8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HwzeQfhaWKZ6AtR3YEvCmW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L88qpNCq9pQovuRtSNMJFm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ez1vf5uyTDz9fDi1Edc7y5"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Yc1DDMPBVRkYs395vobP9V"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1QZWSfJDudYwyykofuS2BK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H3NniSGtrcL3eNHV7usStE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EpTdX2vSzH3xyLdzb4cutj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SJC9nW6LA7vw2q4tf17k3x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFXcGipRusXpHHxkGphpny"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JJzJJujZDDMh9XBzJAWAYP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8aqsZSJfyBeau5yKrHPqbg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FUf7o9A3c3TsJ9xrexrbfC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WyoCHUWb1jXpRCMMCyYqqZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AzkQGk3vs28gr5H9EtLD4X"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Ndc35cJCZgocnf3jVgSbN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-21GrAYxNpiWCFzxWs4vhcE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XXdcHS9o7jtGUDNnbVguVj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8skcoEdnHEKbcNHe4PmwTq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qqcb6m7657r1URTiTcsdcU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6KWTnGb8AeytNfmVT7cYFY"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YDxyehK3dph7ssj4su23GR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MmUkWvbSNXkKNG36Sk5c11"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bd2CQr45qXXtUCXmRy1Ake"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dk6sDAgVqfVQBiDCgi5wmg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8k3vBhbU979o63Azsh7V6X"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R46SBani7fRtd49tiPGZZT"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YEzcvySubSkvuFuAXhBCqj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KjLDvRNhiBn7tMMQ6dS9i3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CmYdfLnJ4xLBVCQpcMFKMh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YTCUjMWzeEVVAXG2etSVPP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PVgCvr6xP2PeFaV69s8ctq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6PuGrpnnKp6HhWGxoHx53M"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WgEZJ5iK2ghu2Vp9ixPQhx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VtvhM6pt2JXjaFsX4LfGd8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9XhYXyszDEF9XYfrEQ5gvW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GdccJEoTUgc8K875DE751y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D9A1CTqnHD18yczX7BLYkV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5cDZbeut7BqYsXvn5NuaUQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T7ppgUkPZiUS7yK6RnrzxJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VKx2kcM64ReLjhabNYRycF"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UDq4umkRvmdkhsjxmcrUmN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1i9auj8DsUv4NpVtaPdQMf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MjrWHMBDB9nZZg8KjsGvzz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4sLoLCW9ABPFgW4AvEZpTN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MA6bfM9eP6NaUPcxf6GRQ1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5uwLAWUVutKFwWwGYTWN1m"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HX6JCoUGoLhgUSEXTx5XWA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CrA9UEbiziG4Sf86K2wXEr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7kZ232W9HYDSx4y9h8FHac"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CTcrRXsoaXWzjyTZaw1iH3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKhohAc1eU8vTYBqMR4wxD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NS7LyUeGW46UHzC2oHeDtB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HqxmsWrMVyNFguCZEWxkdG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4AqrthCZNj23aiaMZtq4yh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZrHLWWeJRqr7sPbcNQZQB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4gxqtFikXkmg4AZM7GF3f8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K8zKUgfqfAcPmWXwiEbtdM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SLTEWKGdhpYwJ2B7iYbFTj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3CbyPYqsXtfYAH9WHfZuwL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5YqWrm56mDKdLcGg3NPvzX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HjqWtLnDsM1ok8oXJZHbGM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RPVTFWfibZ7yqfZYiqjixr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wp3QWXrnYBPdmFSKSByDxt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D37gcmgyQb9t4awyWP8TKb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3W7nCMdmxZkiiccxq8KmdK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BYkKsD86kXdFQEMiFHhsWu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1qVqwSeZoonUXYErS5iEys"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BWbudendgswF3wxqjTuHPn"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7MXHgxAh8Jqbxard6eSwRw"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4LL3d37YmJFWWFwYNBYpbM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NNJF78Mk3kpAYtwmtgqjuq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q7gchmmqsZ2Ce3ffkMgqAt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P5Yfpk1tCBFhGiLjKN8Qu8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3eVwJ99PVvLjgpeh2RHPMq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JDVvhgg2duncaA6ZF8B7Jr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gsjjz4kCQHhdw3fvTstsMv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WaVNboz87xMb3zHaRdyGh5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ESgcJM3F5L8mWRzZQymM4p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KrF5xsmHgWNqY5GBvckDXF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QGspECYsnVGgckxmW2EXCg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ANetmQqvi2f6MgYtuGTEXn"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2XieCbUyC6ch8fSqkqLR7Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T4yJCbknPbHBagHvnAsisJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FpL2mMJuKm1FbZm5C77o2R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DAcegZKwDgv84UmmPg2cnz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bs7JdCrJhuF7cPYoR1YYaZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3NQjrYzGMomvFxZoGQjUS7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PUeH9if5mNqq68yxQ4wRcM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4etgtRWRTZUEoFUdkL6PR3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5kg8kqbQuqZs2LuHyzWByx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SziBSZGDNEVqZe2XEvnz5r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SZ3pffcpWx9Ha4s9xxSLDW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cx5B3uA4ozw7mK6kW3T5rz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AfjsdUU44vmRE8genyPEvU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2198D6rexZD5ZbkNuuefMN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-745Y8PA6sZazqmngX3h7y9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-247WzEcxqegumQHf8q3Vkn"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8GRwzSbyBgd5uy7JSvi7gG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FQF6mEKRUvE2Ps8ACV1h1A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WCq7VNnnu14WFr7FrteXMM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LNbvSzHPLBLR2wpHqA4Mtp"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AWiZ1ZS6HgKZ6tSHF2Afbt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GsrHPqnXv5fGD5SSKXbSyo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CbCxCH2D1iFqvnf6ujqFYC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AQQWsWo5nNLw3zmbp6tEBc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CfGkunjon8bLxqbn384hg5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TKPWHgFZxeyuj54EQfAhNU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-URWAxNNntHhpPs8sHcNbVV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WsVSnwJ65NvxTLfhneAydV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JFTDwVzF1oP27NVUxAjLTb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LWoK9KCNzuBU3VxjxEnftW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ReS5u7UKUd3PrLWvGyWAU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HoQPJCLJjgnqJpLGnHcTpE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BrvsHtpv72YuwvqC2vfehn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2X7VfziEhx2pvGq43VErZK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BWfFzigE1upRPtjqkXw82g"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Xf8q5AF3nFbSL3m7ePQ38"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HraQY8hXSygdETxyzw4MzH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-91QXgzWqR8p7ssN7PVJSHT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-65q5SMDsxHeciqJG9uiS15"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CaGp3dQuMnj4NCkYrShety"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Defqw9c8SYjUpx7PszJ27"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ZsrPN2L6TDb8abgZx9JB5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N94BZJJ64wqSkWUvbKnBkP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LEY4JWpK1uLprfKQnv2R1g"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HNMTF6ATAUnNxDiHFA1skR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9QTaHYA3LXoUKhyReX6Xgg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6nDTHuYwDrRmymDcAqTFpS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pq1wjHGfBGdSpLyxaQDmqz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2MPEbDbv4e9zhG8a2K1jZv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HjWwbdpQVY7yE931H81jM9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EWsx5UhAqkv8YEEbpQhVZY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SbVWV6fLRNMR3mnstffDHe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lq5JWz2wWKEqUpHcWX2HtK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9CiAf3Lk5EWv6kuQrzhqAk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YYYbBjuVNJF4oseqDTyDdq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ho9vHbKNGnSMckcvQfQwto"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6JVBg5KtLSPRCeCg8eTDoL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8QCw1FAM8139wKmvW88qoe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wLrHTV4GT1T9mT2u9asi1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DE8pd9i5e92Rtf3qfYEEBn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6NLDUmLve7KTQpubaPSvCa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T3CGfTb3kQzWAok7cH8aQg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XRG894mHxCeXZgSwKfxEGt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-23nekJhYVzTfmshkWVsMaD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UJFNukBzfZBESuMKeZ7UwU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EDQoMoWfWbRURCy1N6DZTw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32NTshPKWqVNyznPwaYbvF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FxqZ4SYnU8VSqYvnhc2hur"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YAaktW8Bg3TGf4zvMduJhe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-41dgqU6zFmzEjRMY61GQZH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AWQWobji8gg9H9zTw6sP6y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BLe8QqVRF3RTrbUD6vGM6H"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BfaBShT6EfXoxrKsx1zsJ1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7yZkwqbDL6EYM6gopAfPHy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EWtyUTQsgeEwqK5nMbbLhd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4kvnbPoXN7VMbFdbiHGkqu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CeBpj9ofuRsBifrjS5i1y1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFU2JkBuBTeZn7PokT5ETH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SRgnKUr68Fs9ku6oJmnrHe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LxBd1SjGctV2Xk2CZoCeJs"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T9DZna4CV5ohi2dDVnLHhE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sz6uG5GjujmrHiTUDjhpsG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AJCgMNoHHqhDTCqj9AGGgr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9LxhQby35i6qwYg95nho6j"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5xw25KUVkF3r4RkspAE2cn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3oPwhbCnLqNWnjevSiBxKz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFZDrvKxnvok6wqCtqPFKo"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MkUR66FgWefCRSEUpq6tVH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Azst36aaRLwshMgwQwSpuV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1coZVYNvLY8hygo3cPNskG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mezmz5XjXxz2H8qVrWt2dJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-37danvtiGQwZyvfr1RvGLr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1RNDMLAJaLrKCVXwXaHprx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8wtxyRSgQ5wS5oHyDpwRbF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W4jWp9fS9KpVcJ9poxTnFV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8z6PNx6uqopksHdH6TqEa6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5nwqjLPMvnGWKsUwJLRfJC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9yjbiWDoNUEk6Fmaq8CbVt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XYcAtjcuGid23cickHs6kB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-18Ku2K5Wd776ec1oj9Xrf4"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7gkULDnXHE9XQm9ydsgnJP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B16HwnDtNLLMohnuwFoFEz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7bPNsYGZ9vSeeUPETBPQcT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UdVeifMVPXMiNjappxeTuR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2RfKA5xL6txFNzRmoPz7Co"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ni5cktWKkV9jyJnaXE7jos"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A3rzw3bCRk8UKMvXCp1L9s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E237gBKhk99kyG7Kkb1LK8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BemdTddRbunP7fSMpHTk8B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SWgURtYqugtpPRUykXwNyG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5idySEEcScd2EosmJE53VH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQgn8jQZuGHUWG9KNvyqR3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-De8VmdjGnJmHmtw1ULfiQo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J68erHtPfWFtedEia2x642"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M8ZbqXGrKunExZXHy3Ahrw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P9ejKnECvXJpaRRyr4Tuug"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1rh5pbs8XeiueoAEwBTmvR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6aXvHvBDCKAsbjd3SqCNke"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LPQ1frpt6VYkVKEavGSUDr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Tivbv1ctTeii9rCRuEMBX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W9iZA8xSVajWBZLrMWByHf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRS5qvLBaghhMdgDHLejyF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jg33Nnwu1segW1pTdNMdmK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4UatFazWYs34WyyU1BhrBz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RxVhNghe1yJeCLBgtER1XX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3B2Cvpi1NLcYLTtSJSvmJq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LUZMbzppFouVDWP3z52d9p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q5A4gMPhpjyZfX5KtJ8zFs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PRFg37FVFhrpy1BvLW6up4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LFieDxe47KWM5dKgLAtaQv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HzBLgaeBDhpYshos61i3Hd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D2zL5ZQ9pBEW3BiuKwaGvC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9PWhVVCiZGnKMVxcGRUbaT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8KMTbQ2rwEBt4tS9MVc2j5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FgX3oSa6ZxHZHdsxo1tCWx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wxn1wa2a6nRB6xiTsAjaar"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-McduQ8ez7BmRmwVqbMwCyr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gc73QnffEyyStoH3eUzG1D"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vi1L4ZFawcHPqph1xi8qxi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9euL487LnXabgtvKtQhLfb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6zGyUJ2PvbksPSfhTKyHZX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JJdfbbsf4UhJwqNCQyQf2R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qa7UB92agPpwBJueo5KR1P"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VCzXXryMDedy4DoA6pSE6b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8bbdYDzCtrPHjnctWdBCiN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L9Kan5XCd28gzJsu5AF5Y9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HJscMe3z9gStNb53Hdz6vE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CuwK63q2ZDzLk1Q6Y6YeYb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B7YaRLATtN2UrWbjh28d4E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UyoSeFjvoeAkgtWPQvmsvu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SfbwQrZvM7o1hXn8XDQFUs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CDemQEXvPktUdUV6n2xKJR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KgJAQVLSQkmL1fBjumJP95"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BmusPAny29X59EnoQgcvBm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-39c2QGRdAHwaektniB2PJy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FApQxLYZrGrHnEi7Agc7Ks"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W1zNgksfNqRJZo5Wgu2FMC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQh5nmMpcCkNjXkSXEhnbZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9qGsnE3Fu4neTnmhnLR5wg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SVpGgsyqjre3rcmLyLEX2K"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NpPya59gvJCyQxeN76Ce6t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T6KXNJ9k98DtRcqLRius3k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4vrLeMNrzW2GYgD8uEFgDr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NCPzaTcZ65e1thJr1TbuLE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ytoaanSZjXCNPNaH9VfaM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TXCfvznUem4oeVaNRxMMxh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9YwnXca8yfB9vAf7kKAWxY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uq9pir3zmyVHKPXk9VBGMV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EUK9e5Be2x4F3QYYqNrt3B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T8gS4hTCo375phV5ws7uVF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2hUWxzwbNDQWcVniaa4H45"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-955tPSUsBNnEJaVySxzzeU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AHSqwSB9aV81oPNyZbcsUB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TUZCNjySJ7LpR64sbYXZUB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LvTj2gkXYjZj5NRBtYRar7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8DRKRVvp5HqyYnarF6412c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5NCMyLveMdj7in5rFmDtWG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HH6zUGLRV5Qrpo57oRRLhP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4YNHj9JuQJuRGwWAaoE5JV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KLkh9sQoY2MuyetPzssZen"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NrqRFq8X7YMp4Ri65WhQb1"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C5bNgBopYEaCpRjS8p3K9B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K4Wz3QKCm3S8aFJjAxFc9i"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NVzzoheYiv63EzWTZxRdEq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9x2kvtR5YA2yhsvTWcrhU9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dt48dVowPg22YezPVM9aNc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3z1tJt1NqFH3ZSTBWhLgkR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TV3RXmLqCUcD8UyMsVXaaM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QK7RrpKJgHWsyBWrhbTSkF"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-89NhmW9RoHMG4MCGDobpuf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LyywJ9S1kczMm42UXwz9XX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XHnMHUYgpVAp8PQ57dkMRt"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XhM244TYZiATMyW7tEWQwH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PYPaC177BeKMCN2mQzvfNi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-REL3mKN1EVNcadPozeY22L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DkSpGrbmBn5NKmCv7LhaiJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLk2NA7AkYibVyi6vB174G"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MkiV2p5xv3URhJkyGhWa6u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QdKrf3XGhQwUxHiNeVsyid"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U8vRwNcGVy3A56oEGPxkL2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GBkdYVoPjFhmmMryTgrK18"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DbvMdowm1xWu1mZdT343gh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8i5U2sz7ThbmEzYXbDZwzz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9xZB5hHbWuqbRL4YvTbiSu"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9UytERMVbEPAL54kkhDc97"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JrWEfVrXRwnHYYi2etMjh2"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4KRgtEPvnbWsvBQpV48Lrw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DLFyqMCtbuENBYQRHfKhep"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Uzqo74hXnmhqx9tDZ4iaM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QdASfJzCHrLbn93N9bB3a3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WepjR2gLC8UD6SQcqjqX7a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q138xZxGX6JedxCULmPfmE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6gbCF15qLErPzDEFAdzTtx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NUU3cjMVFmTEJUXjTvC9cu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S4ZCQPmnueVhX8zXQV9uCp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3j5T4cRF3JPB3t5xT13mqg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tk7j8AvHbSJpcqWWvMhSuX"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3B3Tx8ApHdqE1DA1XVkf33"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQkhoCUDxZZuVnQJBxAZhs"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3vPUzCQh2ic5xvFWQMa5Z8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CBui2FpPa86uBye5na4Cdt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uh2nD4C2ASTg2T5TUP4Qbc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WSEntMJrfD6Dp4ewmMfUC7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6BSE2Uyvnu8pq5vkDZiqA3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xg9K6Ynhuu61Cf43oCWygx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EyXt7Ziw4szZXc75rTfxpJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HmWThXiq5ucnNrqt2u6mfa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pttt7FTEfs9KD5mYaSjBwA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C6FgMXoNXXE5i8asTX7u7i"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AgFNkCJe8wNtU4TdK3awV9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wg7y9yti4GXitnecB8Qegq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5pDn7RTdNqPYypCduFhCBN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BuzjYxHHapFhimJ3VDFZfb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6UgmiWF8CepZwETqE6uMiv"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ti6k4DkdnrhMymteenyzdm"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PyEWnvE1SBenJ2iuGGe8Ee"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BKFUP8uv8artvSir2Qp1Lp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GnZ4RhPfu611VheTsmf9Eb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MpkhchgHTjmS4ZJuDZnXyp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DMRa1vzqJEyoPmqJuoJo1J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U2Kgp44m8evWWmTRBV7ibY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KgXNjKUqGsHnqxgTXsqWDk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EXuzE69gQKMPvxFHs32Fva"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-41cWzihMf39usqK9SvCJxL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Aw4waJZ3idUPqLsT7sxhDr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JqgsFg22jtzYdDaZHxWEmS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AbyyeE3jNCELBfG2LZAstu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hhr5mZ65n3C6CmFrbNWFnS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WfEVSWiHPm4L4PW6NfTL1n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Focc5C8bBoS9SWGrGsySPc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G2CqMukJDe1aU1y8xyr2tr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NzyDJF6NrGXer9GVw2buvw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7AEth2b2TdDfMvCRFHJDn6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RrPKAEMRz3WjgDwCQDzr8q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WpbwduuyNDTHowfctpG9eY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XVK9pqEy5WBKgsb44uC68G"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ANqd2sx88cRsL6q4qr2QCk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NyHFePoGyMVDycWahSHrkD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T5jTZbWEh71byY2Kgg1cbN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6xSybDvwPSxSfA3dUYmDgf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BNWpbsiCwUg4Lpik3pdGaC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KmwPFZFjTdZiYUJzEbr4vU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Arsnkjm6ccdg26dqp55qsP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JLrUGNm8B8VTHcFqxRxVfB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ju8zdzKAyKjiZD1mVtgWEa"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G1nG7NoCh6fD7RhDureNNV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AHYi7BwBG2VEnDxK91M5Gi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-775tyugwfuGVG8TMUzWwe9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6zCSN6huvgbeF8da9WLnUn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3W95RAt8u4SEaytNaJDEvr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7GunKq1id4X4xp4xp2rxWm"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TVAN3y3HN4D4pQFSW5xQrU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MSaDftuYjqGZfrccJrELaT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XtaW93pb1m5xpdFn4k1v6T"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APXuoD8HLTMjAzUB6oMfBU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-39X5uTm9G3pmyFrpvfvq8j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BwvDLTa5xFMfw71YF94xka"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WNWXsTfBk2qmWU7nn636Yz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JcaXGbqxhRk6xe9qipgiWQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9Mu32XwJu14FkPw9vzKSp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SL2UnSHNTwjaYA9HptjS8z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gwqvmm9twdrWbAESM4gFag"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HVLSX3xKavFUYug1c5f818"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FVqDRezKy8tsnHSLvpSvsQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ywhEMYXrxFmSXqnwSPpfj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6abG4DEdAAWC3szsD1QKLb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CGk1CwT8MNFdhLRxR9wTp8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NVHLtBYrVxF2euKd4Lc7Y1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Drza9ymE9iFT9t3RKxFpzL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ncYGWqQFx9aV5C3Jg4EyM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H98F1WQs3pZaa6SgEEMFJn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YH5KymwxfGRuM2ZMWpKRTG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VnhkhY4US1w5vENyod4tU1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D3GvXwJDR9yu6x6CM7rZVG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3GDrbzXCSCeWrvNyKwJBhC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L5GWZS9xXY4p9Be5JPmVds"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CDrLgb4nzB3HFDDAqwR2Ni"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1QHUdaYEPMUN2V7k7vd2jS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rwg6kzTMcxUe6MtpBJtsDB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3TVcgi9ee6y6VxMVkfcLVF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TdE3oPinVXTiv5kyQUCXXc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LNY4CNXsyxmRAD76DENnFN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1qAKrDCrkBAeB25cdL1MjK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RVVRZeXcphcixYwVEKZQnZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-14rT6Vf2Abc2onaMLhh9wS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BqSH54AYzDvY9uhx7VMMSh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WnVhxzoEWZpFJyuNak6Cg9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MQua4B9gHUHe4pofBXfo6q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1tps2cF9pmM3uvrxVjyUA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1vQuH4FvivzSSam6fTkFWJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NKvsmaGgN95vKoLupeA6Pq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4a9NCNbw2NGNZgwBCpJqtV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XrYEPSJG82ij26DBsCDF2E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2tM9c1yFgokU6YDpwCZoPp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7p5S6H4GHy4bqKiihFYPuY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CYeUzuYbJkJDRRBdbZtF4F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WjnpGPCvrUpr4YfED6U2mw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YNQD4x6k5gSrdvBf2teZyX"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8hcpm1VTnA1K2LZxm1PhFH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EnTscqidWFtYGAYb3CKpYD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JfU9TKrDJJpm2JVMvzXFJ2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5JFMMLe776Ne8qyJz6Lw4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3VMgtKBrdeV5santY85ip6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PLL4vH7bMn8BLKy9euZVgp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LmwGNt3ScTvokqEQnZpDMN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H49FgkZkBJaFmmGYkfJw6V"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2Ndd6YEuMFj2d8oXakhUjt"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HrTPXWT4W6fHxqsEUifPCN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NJxr33c8pR4vFwn6GrYMiv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2geUwVdkpXx6FpuPMyPvH8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KsjUFrYQFCv9WcTAVaRQaA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7kGif9dhWFaaCsvp82D9hn"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KfHYChAwwetXfT1FSWZFVz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B3YgUSsbtSng54s4oznG1a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pv1AJbG2TpkrjsqR8NTvd9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-55i5LoDPuVkoWJ7FBoETRK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PBWDSk4athdmsBEtfHHWmR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQ7RYbN16btgCLtav9LMKp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VkxSZxTG5jXfcPLtrGpjPu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ZPCNmSrt33sFFM83C7t3d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2YuszhYAgdT8osBVeEcpTd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UFGo4CZxSY4yHVBZobfo14"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GB2CUaFQxnnwGZhTwiNhwi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2WqonPAbad967jBqiJ6bXi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9pHnNSbmEGhTB9m1Xe27e9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Np9pGXmXhxWRRkpDPgEuYJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4jeiXVQzSUeFdL8B9zJ85W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1eN7uHDwPrVnB5qqzFkfGz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A5KpdtYryDBxJkXMpoGofv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R4NCZMmeHeTWZXSyS1MAcL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-77VaW5UdEiKiL1SBL46CAs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1d5mDpetvJrCciWZ7ZpFKs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JNcYKcyF8nMBTBm1Bymq9J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JVqLD3H3iLCfKjaPLi1TZv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7W77ceuerpPtXhRxQgKAmh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9R8Sdz4ezad6pijRuJF9NM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VXKY5duCU25MPns4DLabi6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2wScS5bszxxoWAuRwEpuKi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALzgEpp1zzVd3iKv2GJPjB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsuF5jB6fahNLQ9xVqMDyh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EyRd9Geck6RSbzY1d61oVf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WmFJ3K5NtvXPdyTFCeVLVo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SaPcJV9YLFhKpkiDjtt7Gt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8rfJnhXC76p79BAoUbR5JX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KYxNLEWvN52L4Jem4rhtWj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7dBhnBq2p3kxuRGfaVjrpL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TaNEN7LxjbK1QJ2bnFYUcr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ao2iH8WwNVgkf5KXtNgHN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rm4vza1XSoJyWay1L7d7Mq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wv34RwcaMm2JUJMqGcb1Di"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBkviau5SfKiCvYXNX4Hba"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JhujdD4i1MUnBYKjjBbwLR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BJduRD21Whsb5ZUdxbMEXe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-76C8AJAuZTZ4C6X3Z5fmHd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VW6nAcufEPjghahYth2sPk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3R8GQtnshbjwLeswfp4N2C"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3weh2m3pBSx15D2g53vPzV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DQkyavjTEW1PF1xYurwwPh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q29bQKX6m3qA2ML1W34gv3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2H8srJkuj5spxByMm1jgTz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C85Gg25tp4mQxgDRLqkFg9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8iZmbbrPzGmU2oDeWjX3x4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GrfCxy4uk7wrdsiG1UDBti"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QvPjAWGnWdszRg9zVnBSVz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PqEs36ePmHyZkjBcuVUwQv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jh2RBX2VExeZmfRhUMgknf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KkpELBi2tE6BJV31xjxQqL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EeAeBPjng8kk3uTGMW1Mbq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6wixcuDMpM1wvsBJoAZbMU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UnMNkZDTJqSXNR2gBZVb3K"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PtWmHUCDh8YghNMmmnHrEm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8K9CZHfcek2M42nPWdNd7U"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5m3Cxk6SXdhnGguL8Zge95"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-An8ds55Tqx1yb9qHeEL4Zd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VW7m2CnE7otUpuzdsrW491"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KHdccYoBTAWNvAKMBgETPy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6VtGT8S7gNtvBbiZBXufFm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RFCGwF557GkFuReLxYrYdz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RucXy9bnDYZvqVQ6ddLVZc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ACXUsbFxYGXFKcLeWZwPKF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9bey4er3CTX79MaGxMYmWh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TAvheS3wmQZJuSRoGyq4F9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BEyHNhaTHgGHzYrwZhSVLu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9VZVpZT1Jk2g83iNcTH3yY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8sJ8VB2XUM4Y2jA1k8Vcbe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CNU9zNTf8FQymfs97cMeb5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7JuRKCytBgyYK869AauTMs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RZ9mWhLHekfJSLJP1HZjtF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HnomoDwphaRs3qhn7kdXX2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SGEfMqZVSSrPyW9QXhd3wr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SEvu5jKpiFSN5ZmtDkcaJE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CzEp9x7wzajKvqKbCZDq9o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R2wpKdYRjcvbuSMFD3VYNP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VU5qQV5B9TydwHeg3ByvLb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1rwCkmE2GxUXj3P2hSEyZa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Xsx2azugU56wymoLnAPTS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BErsdViH5nR5iKh7hgKCzp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MLAs2SXsFzwJorHReJytJQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wpi315CWKZ9dbKP3CfK9Js"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q1WnAiVwUEHKPPQGcNgbq8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DM8RgDxdiLvvkrjzpQShbb"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2DSWheJ6WqBeF6xzmN2gn5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U7teE9PL64JAJ8JXDjSa2H"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jgvn5jvuZJrcRGirJa9BKS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dv8u6V2V8y4xfY7fuZVcGp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EXd6WThpQmYwFqY8K2n1Vv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CAkT5teAFJ4w8fx9H15A7e"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AdGVbBQksXoP8EDmhHmL6E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UZeyBw6KfD95NsDhLXEodB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4qm1EuSJiiuDSQALKvvwAr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WEMUNnjBA98Ziu2JD5PLrF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3enyDmjQ4SRhxdXaqEcGAa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QXxgzVUhsiRRECRxYbeH41"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RSRNQMX96a3YtULpDbEPZo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4L1evCg8GPN3eFURe1fzKN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HscFCSWeCJh25gFvWXXqKe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Wmtpn2ELBBJoHnGU3276e"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DNtBPyLrJStAAZK9Rtcx6f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E8aVGFCQzq1ZKFgrh5Q7SX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XiePNZcFgG6bDsq5tb5DXS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BL1La2Gk1GAt3wmRKu8ZDy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VcHheHfpkG4MrMZT4Fo315"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4wTzM9C77s4AMHGo2jJSKi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7xTPYmZfrhjK5r8Kww1pEk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N4whKvMcJ4TyngHZyCHJA4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3xnYLEStinTviUPNp4tRmy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9B91HBYdnxwPd6VZSFvdDD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2p7HApncQauvGiHrQ8R6q2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XsnkRQnbLXqHcvwok9Tk9U"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LaJahHh2aVJXRAje6ydpfK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YHfawhmJ9zzZr2KL2g2v9G"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ei7JZF44eU3suMpRsKxGAE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QbFrS9NCjrqBpWsZABCLkA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsDu8omFJ1H8kww8sfrbZh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UoPfYq4Ng8Ek5271zRvyB8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T82JDzohDdmS3nQJwrLkap"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DZojZfqmBjkEmMPMcyNNCy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SzVzztsSWF2zc4dWqJHzRC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBuYZDmtzqqChiwEnviN94"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3YPaFpRxfANJLvcLFhhCvr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6CVashfLfTzDWxptqsoMAa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8dmpTUdChR96AwQA8iBgn9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PWTSzygQekfuoG5mk9eubr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZBKw83acLVK92wJyF6JP8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4vY4scMGf2LW2YdBYbrU2q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NGun1rT5yn6SFfkN8AnNDb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SrdWtrZdjWNxVSmV7xt6s7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CwMkotFEbGUnEz6hk2F6KP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1V2wvDGXawSi3Fq5WiSiKp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HZEwjAK8tAtYVjHoqKRiR9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ek7KcXh5gsdZgyhJAPKRm8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y23QUmy58zuCARroCp6KMt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XL95UPsVe3xSni3A5nfJ8L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3PJowdBQjSnKYg6MFo3nRZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hwx9CNCUcsDnEPCq2uruFr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ED6YKGqxWRD5RgxCzoV4vt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AFJYb7EDfUSkh82DHjQCm1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ge3Djesw6sJ2sDeK89oiBk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5gz46rZJuQZMFuXDk2Efho"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MYFXWf5dVV1qJrihpozcYC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ns1X6K2iCo6PJJc2J2tR4Q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UVvaRGnRqWTp9VasiHxFsX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MoRgCPDHkgAae55th3iGyA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W3JfEqwXhJpr5RAYGyMq2P"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MAyNRajcSyZMYPD2xVt3wA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W1zAGS518qdJ5Dev9SSAgm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-89yaroUDsspXaH2ged1eFA"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RoJzqsvoEUcQHCKfLehQL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TSciNrjTrfNuuk4hyPoGfj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U126w7ajCowMF6wHX1qzn1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2GxggKEaxfNi2zjAYNefC2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JXwrhMcBsQSi1paFkMZggq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-57UZCA79x72c5hsL6z4A3G"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ELD2djTq76vmMcoXHiSQ1e"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FxwJdjTiCeG5S17Z56n8mA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KXABVtpmLmPfTxpuYnChtx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GDKhBhGy4zYytx4bWqyipr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HBfiUwBzBxuKaz42rWVB6t"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ajvui3MoPxFdWgjefGXdGh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EggrumZy1F5GqXesnM72c3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HAiTkcWcHwUKtpunqdc35D"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FP1UqkaWNvKP8ta14Vmvg8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2HZLrsGcBiHaynPyBZPiez"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9dK3TQytQhBvXxc5rEVnQr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-54TBMg2griByPJtMr7E4FL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HhLPzAjkmnWJ6dkHnjKAR4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G5QxYZkVZDWkVYXXbdYXzv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FHDLZSA3darA3iNUKhKKzU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RaNJ9VBQMLrpnapCueZ7pY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mx4fCce2tYbwy3NRtdvJoL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7VZMqFCff5WU1jAMcB9c7C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J7w2HaW3MKzMAPjhnvHXkQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y9ShFxYLNChSik1Nmc6zyc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5SmXgueUJuqvtHLAVhj8F1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-245tGxPsTt73Er1146nidF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-84MojHDtaRxaLdwc3KRj9S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32LVRomh8E6fgPPx6VypDD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xx3zCDSgMbZuc7gyaWL4Vb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UA95GE7UvED1VLHhiwqThv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6RvWTjFb5isGwNsppayKKk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B8Mk2Vd2G2xcxRXZaGteqq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DJ4A6yEqJY1sndH4R8gjJ2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SHcxWy5yfJnUugS6GdjpuJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKLkkLRbKwzW1TKYw7Myz2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-46PPELm4LyNfKdnRKyssuo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5VciK3m9uWbq6rAXLWmQ4W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2U23nqXB2Hr6WDd2rpi4MU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NcMoqg2ygfUj4rGq9SchLY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GbNx7B8U3sqtb3uYviZgoJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2DYpby69Prz4yg26BLbX1b"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7JeJ3HrUPoDj3QdRS7sV8u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AFL3vHeYiFeDVAznnkPbAb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J72pmWMez1sHiiKMR4NERR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J8tJLbkXXk3WX8Cyv7zjGx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JrBXpnFhafddCAMMb1cRbe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NGrQm7ktENiASMmC6f5gUY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C5JyH7c8vck4ovBHcVsAXp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5rFE9EVUmhhBB1ekE1jPtQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1R7vMZVcQ2BEYma9auLjJj"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-55bgfeqBG8APY1niNn5VMh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NQe1q1ZEBHtzjHXfGKv3pi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BehNQQUjpQSJMPBzf1bWDK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A3cHPjEvS6QhqtyEXEmDHX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NJFxbuegjL36Jc5rYUktXP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M14b7RTB9wjvyJnepUC3vy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K226xV8UGaurxF9XHuwhNU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LhTzf5tdRsb1J6GcnBj4zS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VvoPn3oXQkupfVP4HEmor4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HJ1cwDtcsgZkSCqRrNPmRh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RE7iFpDJD83HkxJVWYAwzi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gw651PNJS26L8ULKWLm2AT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U4SiYj4U52K3zVbmQn1eCd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LkMZd7d2mJnpym6xkq3GyG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LAmjConnidi4kypMCH5o6m"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Df9jLeTnGmfaMbS8Vw3VaC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JYZqvnP8S8iUjK4nfbbRaF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uo3PnWNSFzFzeA3iiBQWCT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uzb48GamiTmn3ZJcKtnUer"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RnkkjMSqNmefh7nvQbvJyx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vsw2WYxr5g9bkmkNPHMdFC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QobPgmLiHXZBm1ycUXctXj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ui7Wz1d6k6cbGNEnD8ey77"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-52RqkjR3hwomXdsFNXUFSn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PeXSHJ29ttgEvguR1VCKbu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CxFzkrZM6xMnJmpiE56DMK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SWv2k7jdF39JApsJoutCna"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J9dA3xaNskSBomNm3iEFs1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mfw8CVprT7Y9kRzudwkf4r"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D45RpjEt87gz2kxTKtkbYx"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BzUGGAoH6ShTrv3jEoa2KV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4eBBsusRJ4KePY2ztuXUBR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-75hGtnAMe8Ng3LJuP4Wze3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JZVtVknvqb7SrgipVW5FkD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G5kt5sdAGnJ1yafyU3TAzo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Rgu8ps3jTD7mYKaVJDNwi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bts5nHEsKo9YrqPaLAwVvP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WUR1sXanJeW5igokZ41gJR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GzGsyU5Wi4tVdQT5hLTcQd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H2wYxyTBRYcP1hk38NGoL7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4JDJAhxdfr1c2rmjfaiba1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HMNvEPqcHQwPq7K7rWxHqh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G1msmov5ibEuhhvQoP5Qci"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JNHSkbvj1QNoojspNu4XFR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5JDGrEDJMbqZ4BoP1vidiC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XtriyakzEnoYmCAFZd2mpE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WpRodvmxTeSnqUEZHeu5E3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KwFdTRgCRpftwVKwm8fmsm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ApSvwyu4WufZUhBzH9qnM8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5qitgnambB7M8Pn82vtiLf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F9wqmnPn5YnJi9UT95Pmw4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ap7EPqrP1DPzTDqLNFmK5h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UeDEmdN35KLQJ9xA36wxdK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9qHV2wGYGCvwUzTZntsUxJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nrfn7DskLFDTdg5TsGeDhB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6dWdqvTWxFjCdkUzwWvneM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HAfTzKErFfZ6tDdPiKGJ2Y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-96u3MrivMAU8NyB1JLKVDx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8PUscf4foEYM9KpLFAWyhb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9psZwqg4MudZYQ7QtoZBYJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2P7gcZEunod8CK6tTNA3BT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9H8YcTTQesQGKNWoLBMPEC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4nVSd3tRAuipkpGXVQdAov"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ZVec6StZoxvf1zed2negU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGa4wrSPRHB7XcPBKHHc1K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VhQn1oCeGpxH5HWKZEnvTL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KHtAxHk6XUFTKLRDr2x3Pa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KMfgi3JUwgB12u5Q5ug8E3"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4XSb277i4VJYYcqNAhQpr1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QqcycCcMs4PtfYDCrKPMVH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JrVv9EPYEVXFucrtsD2wkm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UwjeeRWKAtFn6Eajxh6JTE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7LCMH52YTXsPrSXK6Pvo5x"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DanC1fUs2wJvYSnJLFJqeL"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5TihgjkJQmxE5WR3H2mJgP"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gc3fUhmnRYpUbU1K679cRn"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LXQDLgrAeSEKXyQkEmyasE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MjdRxACDoUq8e8FjAmYbLG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HCPASuxUwqh2LQw7ELTSXA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5kNxDaLuyNJgTTg1nxxTdG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4tMkXhSo49uB45cHttmG1d"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QuiApQfkvwyT2kbH1svhw3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3qkZojSVqD9Mwypukid81w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4tAHe9YrLUsZhcutXe4YRh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mvub62emdogya9739pi1QW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBjpBKgJ7jCYDMaLi1kPkT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LE4ZqcRgi4qwMKePPWKN9r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8YL3dqq4bH51srVFsC9xpb"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R87noJHFfNdFeQ7CbFs4PK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XNHJ6iH7UiryuftPHy9wyg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8JF2tRKGFDhCVno79qjvdH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AaKK68uoyxQsCJ7wh8nHxg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ck17V4oVWhkmxT9ccHKTw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1xTf3HgVTHuvG94tU5NaNX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3sfWnxE1f2e3WsTbphhAsg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SEDZDDBd6NoZNsawFR47Tg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dx3m8ev1cTc6cy13sUqD9b"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SpWYre3dKY3CdzwzXd6LBi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CtquKRh6V96ungPRAUfrMF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BDs9qPoc62Qp6KVa39deXs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JGK5uUwoPA4EfGTewUUyrn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BbLGuuaGT8tGb3tCCZm1qX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BE7xF9xyokoXyexZAr4MVJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EbSqcRX3haNAkWhJ1k1gBo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GqQGJ4o3qK5kVyUuGKF13G"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3BTKjDvcRKgHQERsh2unNa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Upi5iTHzbEZbZ9obxcLCd4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fr8jNJnw5QzFDvesmhrSd8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1qaQSEoZ6TEKj3MZEgdpX7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lynnjm4hZKCeTrFALKwuex"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RsKpyUcrnNCnJuuDPWUmZC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1MTJwwjKgomotWo1aPmtWY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZXfot2Jyy2cvWurzoxEFv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rw3xMdLAosADvaKWBxNkpP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QgLuuNnJWT7DnWX7FkmNQW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FF3Ry5wFy5vgmd47NpweoQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NRjGnTmPxe1DNb9kYSwmnP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qexuc9Yb4LqMNTQdVxg239"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VmJBqNhw9BorhjhSEj5pxS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2WBkdSgFwfcepXhUPGjBek"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2MNwv5s2ad3avBqTC9RrLM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FTL8HsGLcPV69dmjEdTvQq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-54zfnBPcUf81ZzdKxZwn3t"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VbCLcBZvwkU68pF8jN1jNd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FNKGVikPTztG8mRsfcVpYp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WxCTboNESN7pjphNCTuxaS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lzttz8QcSxHSzh9GXB1uqY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3v22Ugcmsf5Ubyqfu9HZ8j"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3rv5GE7u5exXAruTMk5aNV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5FyHTimhkqr7Xvr3cm7kAS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nd4iNiLQbQbN3UR8v3dizX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-En2H2FZSKew1rvufGLYbxQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BrFYyxvngtAZyMw18z1Xgj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TnnUaqL9gNC1v9pfzprncL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J3CVvoePQXgNoEbs73DMR8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5XU9Fj4MmYtrJx9tTPc9J4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9dzqWpjTX2xkwJF47nQVX7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HcGx3cDb4Wpac7dsULJgmW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7iBZXHL8D8QHNiYxvuzzFq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5kJBu8hWMJvbNPrmcBEApT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QKktEa3oS8Wxc326HruvJL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3VsRgLY7TqvRj4EaS149su"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BUwjSCLepqNSiULP1C8wjY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DYrTtiDS2V7VJ1Shx7saHT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5EYrEiBM8HfKNrQmeJwx6g"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFvCWNVmm19znh1Q2hYfFb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EAXBnuTPFtGXSy6gvmn5SY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-68Cg3GDBVnYLaLpkwjHoBU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Be5S5QRZhhKbC4q7dBSKWo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lt7x1qiXW4vSDZAxBQCMBP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S9fjVFL7SCPyR1tY69a8xd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NjwhRT2aWznevvjLNw4UA4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xa4E39aqbVfvtLsTqLJk4h"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-86SKjyNNhB8WFei5ERKBSJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NyWFziCCJo7RzN3Sosksqa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1AmF2tY83QHa46xufe23sU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CFsPKtkwWMTGKmC7wMjgMN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-McBVnWQDTxpPRySqZnfpKX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P4gyYH6bQs5gM9YkdzagKB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LH89WYM696j7XG3haVUL51"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tc8KhQkwHCCCpbZnYYtm8p"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PeV4mhCvB7UnhBD5i3c82j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hp798qDhCq5xZwy7XNmqEh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BkSoXAfPKhKVBmk4Z6zCNw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AWAYmbyvbRP5Qk1ASgmCoQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PLRgFKRMhKnRwfZGysf41r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9d9NDC657GwGUFdKDMCPes"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6fVE8RaSAZaoy4ZNbMVAY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CHP4QeRwF4aM2eFuZXavSu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X3UBaBd83NXPte8cABQALy"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AZRxpUYZwUAwNoJ1tEsZ7J"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6AccLteneJW4gioyzCBSnx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1XUeQSLrLhDrJETGbUZVyT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WJKgy4oBKq9r2KFRM6ugNR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-16rLzsmH7vejEa2Eam11bu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KAHjXkdjiAzB2eeLkWy9Ze"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QPvtCWMJ4FnyD73PXBvxm3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M2WHZ6YNcBw13YBVucj1AY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tx1fV4MJpGBVVjujGxnE6v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TYFTnY9JUuAB8ngjnbynny"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HFurm8s1YFKGo1hpRXu5Yy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1LHfsp6mhrCVDTWQhn1tAx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GdK1tecRhxsovGzQ9anHoj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VwabmVdbfidjhEowvfi9QW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vr9TqE6X692FULg1jnhn1A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XLVtsWRoWczKy4W7yPHSQq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PWxU97gGWRU7WSLN1FHB8f"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1JfVf9AwXHaUTSrf2hWeCa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RmicHZj8jFEypSkEvyYeaZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WyaYJZnKXq77rFL5Vhsvve"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9wVS7p1JpgY26g6CRo47ew"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5BxrPAu7zGWfCunF6Z5YPs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wvwa6T3Ubma2N35T2Es2jc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hr4nzAPK95wajuFxhRcPg2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NNGLiLbJrvzKe1Qtr6oFwC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BajuXcbJ7k8PcfZknbG2RS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AnMp3Gyhy9bPCCdwJobuiH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KR1WzxdL1X3uucPoQEn5Cd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Egz1RAhxWURRYSj8VXzaSU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PPactyFbwPfSfHtevwyEhM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1gwQSG2heYPMdBYeH9zrgW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LM2S6NBxwDctBN4wNGBhsw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7PTgVdiEMuggqPcu8unqZi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ZGPGsxjyNr7hA5aoq68aq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MKJBmJDMAGujirGJJjCpSy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BLtmroo8o2XDCekYZS7x6E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JvjQpDzuWoTBVmM7WoiyVF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1tuHgvFrYkczAjiYcK48Mt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WtLuQUF9Ht2mNWwMTEDsCg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DjJBjDTDvtjANHBz3sQkQU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5feLD62xGVWJ93Q2fiWSKo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsZeWUZuJ1MWK1qkibU6y1"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T8pUX6UxQESES6h5gEjrXM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JgmcAJDDB5kevyiJ9LTbNz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NBXGmcVLNr3sMyE42mj9U3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1mnAQuuccWk7EcY2b1fjgz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-56eFB6MYH7BozzyeqcUjec"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FzGvG9cDriQ6zG8unPqK8D"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3MCfYPXRofNMq7DqZJyroe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E3UaGYWBC4BKrd4dSchnCH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BEsdp1paGXPoz5oYgmVcFq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BYtwPnAVfzN3E6PcoP8ei2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W4s8bVwPJub8kCW1dyreLv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CsjFReS2TTpJqbHnh63SCS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UkCHpLm79HTpoBjDq5DZYy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RHiZcuZM1HaxXMrq8VSPnJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4s8VgEuQKLATQpc9g8DvKR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-53kUZ6cTbyUmHRZoyg19oA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ApmXjw6r2irpjKCeF1QN3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-STrAbRLF8bvmF2YdEQf7Qc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Moxe6hqPGrams6YZnzYHNe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KyUjPCoxqDzq68VgFw7uzi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ny2qRf7d1quvVCk1d9zwa7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nvi64LCXpBR2uVshQy88aJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wgm2jGncAeA4WTSw2nddY5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7BnXDrPA8haAy1VkeyPehN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KrVk292D79PBXCFAYGgXcT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EGaQwN6KxGGNo5VXXBddPW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Q3ucDvQwayYbMcTYMnUve"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7MZMr4Mh62MwezWpfLjWiH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-95rUWob2yBqHtYjKA2ZKjQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uij3EuxkLF3tXSiFPFypXw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8N2JibWWnRFj2iJowa3c6F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FmrLq3HTrHRQyx84MpJKxW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JgbryJmj6gJXajNwEkNKoK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DR2633EC92KnMBQhpbmb4J"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QBZCXPD2ZQ6z26DeHPLtmV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Scwm9vQkaVXUyue1tntTH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VRuv4zBY9vgketXSjcwWAj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WUijCwFmqTCSCeVj4NQfuc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UwzdHZ7q9NiUnW3rT7Czme"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GTRXmauvJD1zwgZ2BfzHg6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AVMP7XkUoKLdNFez1baCaT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-11Nj7s6fLFfBuvxds3k28o"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Svcx6qeNY4hbgXiowGhoEZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1mpjaNHemD2ps2HPXAHfpC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MYktHpBc7Jkvh1pUEFX1xR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6fVbXbwHyE9bPdUbpHSpjD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lz71SR2aHf6oMRxq9Zapab"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G7DV6XG4CJoTer4FfndVm1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QVGbiwYRsid2yiuafA4NAE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KR9Yd7AMFC735sbKpfYzhv"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V3DbjWsrQYj121qUvKSow9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T6Kkh268v3dozP4uWkCXtN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G743w1wfivHFTYKB9dzkKT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8NbZ5vgHSpZN5UqzGsNDge"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KyqJzmvM5c1x2pWP6NJQia"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKLNVgb9DaeQmYDvj13K9X"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K614XJzuz2FuqKCAJj358d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Ubgrn3VyioZ26sbjF7X2D"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TQJvLKRaHsebFPQSxqnG8y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CDRihxTBfWTVX42mw9FDL9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mash5hW8WzQEromBHJdqP3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-McjC83buUi5iaSoJgfcZkS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MnHAwfWHX4tdcuYXx9QHee"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NowMRXgqV5pk8susdNeRS8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sm2vAcMGPFhEHYxzXVdaQ5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TwFUr7qH82QCcYHkkDzmCG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BhmwE2drmNCVTKqYghZdgo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-97i3EftSZR4p5415i3Vsjq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DZnyLhpzRgxFhqd5p2356r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KDkcRemXcG5syQmvF92Fmq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RjwXqYbLgfHDzoTuT9KuCP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XLMkc8dkCnfAb8dSK3kkkQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CqjdNFCD4vaskGnaGY31dq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PjRZLe2Rwf1esb8RBdiUS7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5YpAtmGAxNqp3wpWSu4Hk8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1sJvnynGMMZaqyfdN5j3ai"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bo114NGrJQeiJ88vQ8G5cS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-544ksCk3JEZoJsmDfo1cu6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FyEdk261iPm9Yy6N6iqCwr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6vzXo3DGASStJepUrCmY5z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-REXF5yFGnjftbQxNCMNDfe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HYoJV6cexAR4GKgg8wu3Wx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sipc6MdxMrn1cbwigg3gjU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DHxwH6qw8vezSsxHL4BqXW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BinTB8VeMKuQUjGgWK5TpW"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KbnafFb9jaqrd3x6Y6ca6N"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EB2fi1ZGYTFwYVYJEUY528"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8jDqVf3itLNGmc3p2MvP4Y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ekz7feNtXWv3iAT2garMju"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8W8BxMNxkTaFyaagcWD5SN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xa1pKXdTJTF2YVtSA16Fjr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WKwFEikQ2yYEayG6Hsb4iK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G59WM3J355riakZPpPSbVT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XALxxBinRYv2KiYXKcNjJ5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFpKuU4TGLyDB3XFTPjjtH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BxQkncFd1LyNXFkVEtTgLB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JiqwCFK3ncZ8gBkh8uRkke"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XWKnMG5W9icgLCvcUBzvLW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HkjzkF4Fj2hRg6JD3MPtWn"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JWidhGP4XPwhNNCfqq8Auh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QE64vJkZyZNBGsQuKzbUj1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L7rxJYerNM73ENG88kavcs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FmUwXBaFZrM7eRpzJk52Vh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5dG13PVjheJkfweHdGbHiW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RX5bg8AqvbyCcKGYvnsbt8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3rZGxsX9Rrcm9f7Y2xQV3n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3w1vaV9kobbTniRMHeLVSS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TTxd35BwiSQU6pgieXHh3T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wmcGh6ywkq7Y3CUmm1VFM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1k6Jrays4duPM3mRG8LJv4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-448oDd6FqyPsUbcMNQD91v"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VU3zitx7bmZarvHKfx3sdq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C9yHAPimtsVqWgvPnJLQYH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MyVFV73ky8z7EVRqMZEZxv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VVrfd9NT3nPLnGFnLq1E88"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TxkcCR4qvLrtT93N2Lftvc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V5xQ8B1zcbH6RXaDegWH8E"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Xx4EwPXrB8aaZDgfWdQsq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mk1BcARrJgi9tYRpnPoRWw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Gu4DG3idmJD1ukayrZEGV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LX62DZ3sK7dr4WJQn15wiY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EQQMbjRFPdpipx9799AbSn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CzrWxDDUzy9UKy7iiS7th9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AvLdGCxnQGngJBs99zYmei"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-292wkEEvhzaMyabioNyLFo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4z8EQhgGxaLrgurFZfuWdk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RYZmRbhze1YfPNVfwTvUsp"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQrF5dQkdXuVvZA4VxnAtQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HLVPUgWA1cnv4znxLjNaTn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ge44sDetwvmgGGoJtZrJSa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5a2SaNymUrvqiPPxH5oPdS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-63meYm6aFUkmJscKeWyZrL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1yz4QLusMBQS4X7jCDWwwL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32iFesFCyhR1Ctzzio2Te9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VrvtZVodv5SjNbnaBhAGWa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GUaLckFc8LxA1TdZuQSUBR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LtXTJ8aFgKCf3VvEwWqjby"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dy4D7Arr5MzD758FxoCWwU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1PHub3LXR5LvnYAjFGS7oJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FsrLhRpPTqJzUekPEDisrm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-385mHiGq4hNHk6EXyscGvt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8CWD2heUHWbryfEjbAYGey"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AhBLW3GGFv13rzGaxHRKPE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K7f3djXcBgDZH3zrfpfJkr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E7tJEinhgZbpDtRzC8y6Ag"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RtP1f2mCEFmCnwS7DiRPuS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TAStpbWh9Dty3kM6tJpUtn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JW5LhbdVGJUK3dvMk68aUG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ab43jUTFs9QbBDexeqRdy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RL6vuPPQZ5mHAQuFjqqmJj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XKqxCqUgnatazacg7Bzzcy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6UVxPhB9YtVHni9uiybsbh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-74eEGXDKhFq384NRiXHdSA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JnGLjh8Bg8ZWqehEbjV7cU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GU1d7KGAqdrufvogT42aKx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U6mG2PfkutDaiFGeVucTsz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qv8earhNeWa3Z6QnRKbjyq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ZsBUzPvpDBNKY1FF2Vzv5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kneo4WMG6CB8Hy3g7hG7aK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PwnRRCPZYJXP72u3eDdA2N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AYiNLcjRcH3apYM7RsiPR4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-84ks1uXQxyb29gfUom1Qpm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9owLXjqgsHP7V42QQ9vbz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Avr5MyFv5MgSdPhPYTcdea"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M2ZBEfbAdSBWqEVvU7acmc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7qc441fEYyVP1ZGLkFkL2k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SntMURTqTCoVUTwa83KuSJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SgstonkpGz4oeCs4CXkrMV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RTN2q8EDC4TvdEUMjQLXFt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S3UUwKuhrcC8L8FTHW3FN8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFqkqjoMYdyt4r4oZMPqZX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9SwwwZkNYf2PWnSH4zSFTg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HEgU34okNhPCES9VkLzTjp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VyPLspi5aBHfBfFxmwMH24"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-THhNS8W3Afky9qZTrZWoZ9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9iVJr6P7sptRYcumTWGFZ7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2BTtvYURFQzJmCRR2u1HWi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FXkaK5cJ3QFicujHoeVXqg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-55Uk3KYknxU6Wp3B98SjL3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XKnDDD42WMhZaZCq5KNgic"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6uFftycS5cz3uX94gDUvQD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JQPdhSeZCpQdGYq3itFsWY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ch19Trj7XwQ3n2fpFahSYg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2KTwtDWpDdg3Ld8roXHiwC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8rRb6LJJCovgVJutUumium"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KTA2ZbUrRoUnG7ZJC2ukPP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3HUh8tWxWyT1Q39nKemULC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-846kGoyc13UFM6PNBhdjNP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Z1BNYPWoAFgCmSH7nLgg6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6CfHzd3NHteUZqJXkvYKre"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-86tiGyTEobos4w8LGwK67u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EExstce67nUC6cps3CzNjs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rzs4h2pWtcU94bcfSAa7HQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TQQkptouBNE85TtDEwsre3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BEjoN8TCqQFvHh7387x8iE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDno5sVhgd3Sf7bGAUbSKw"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FbPiq1RKaoG1u2aA4NTb5p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4n6ELxYFcYUxR6ZfNx3bWX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UGMFS2AkxHtSAKephtVLer"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5sYVsVnQsU2yDhnA9pHtnt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Z21aq5FwpomjcZio8D5or"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5jPHiZXapuTRbwQd6UB5pC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JXofYL4ncdTd6Q4rnUBDYC"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8AKVPTU5NtB7Q3NfAofXgU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XEFPQox5yeBTxuPGRVExXr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFBL8n2FWAuimfFwSGZuHn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SoyT9ZjoHgbkabMWcghh6N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1ZyfEEM8sJnHhGGSMBWXN4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CddXVkP6XvkpQGiguXrSLM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M3ZaHhhyQmEHdtCqYDtKM6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7PhRJG9ehc4Z8ffDDi7PpD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XbvprS4v7L4a2oSvegFdPg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2xqnCP1osXpHg7xaFY7zTZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BDt2QXDxHKczewvLrv7VGA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Z9oCYi4h65TPoFuKJfbLB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HRx362ZH6MjgsJQSgdZn7n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Q1GXBq41gzUiibgbrCUiF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DLVMDQyFvWAuGihzP32fFn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LpQGzUqy4o4C3t8kF2mJmo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KeeCF3FenctFKNNoGjygdH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7z4CUhKJksvBK4KChvEEuv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V17uPaikDyEqu1N3MD534b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DySupFqSeWbmumvJ1g5J9i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NLGvQHK211a2KTqL6Z8cA9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N87GuLcHzjTau3jyCB1Prh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LXpVxQBV4EtnnA5CkMQ2ys"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JJgVdVSWfABUMvRDokNuux"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H6j8tUDLXZYyUNPxAsYmEE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KZA8wsyR1Vfhp8xA9LCzDx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8gQma2ahP6Yvsp4iWMb2D2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGFsgq2NmLbTzshdT92Ukb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5hPEg6Mu97c4E7PpNbYR1H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8yEXcZS1KDtZveCisFvgNH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tm5JyceQXQVP9whtstw6w5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K6hTV7NqB6WfVV9mAVTAen"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LpZYAyTXxAPLDkuT41oWjG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RiguGd4CkPictHTahPdt7D"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RNTcEPu9WEsAjwGStCv3ro"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LSF6hJDBvM9YYfA9zee3dK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S1k55MGBX7P8QPQjdin9Ji"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VeBQGbLm7Az5L2s56xZXa4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ei7Uo8jwnnKbGVQGQ6DDc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Gy8cMWuUEhGXUZ7iCDW51"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F2VBekV9TWPdsuiCkuQjSD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3yyNpXUbJ8sZFSn67hwu5G"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QBrDYz5ELcoZjLHowbT7jm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLztBbSVnNMac4DRyMWRMN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Je8sGbP2LY23iBRBEDuhZ7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4X13spZMdyaiPskFPcgbYF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3GCNvBHS2oDeHYyiLk67uq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TNp8gbmkub6nnKAhRVZZpc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QNFRENdQEt29FexnYbLNAH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6T6SiyYoWm4koiFLPxHNqb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-81VuMAMGpL9ZNqVNLowkYg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UYp8X9dpHpQsQKrFZv8S8P"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JECZnW38cKnNPbYofyDhqn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Km2Rf5YNw9oTAQQeb1Jv2Q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7mPE7YnNWW4btgNB6uqESF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NbH9rimds4Sa6JRYSGhDWY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3hrssxbAqRZaLG6J1hXeVr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JRyar5zc9sBCJLiQoCCiDr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FdQSQAuCuDVX8VX4rzuBn9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKuExzAgzQEPDGKstmGv78"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J3dUAZserdp9cdiT6HnT6c"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MM7cay1AC26n3RfHjWZ59b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4MptNaLHi1Ti1e2ESXHsra"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WZv9UX6RHgyFRxY79CGhcN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AmXLm3UAKgZP3k4rHFjRnh"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LSLHn33BiRZH5SPZYZocMV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-63TxCMcuVZ8J1Lt1g5RhS8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HckbJM7kVC1ZB6QVaXzF2N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BkRDymEQJxgF29zbhyeKh3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-794pBa2XGtnnDy6hNuodQx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xs6xjkeX4MDYhvVJtP6jWL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DyBxGy6662tEQ4WN9BZ9EU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VUG1NPyibeGPqkNQcB9W8M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J3ZRwmafqJbpgnmBtsrebo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1YaiT6QAAVMihzQFwjTNRv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BqXMUu7BFBKRZQisPCG8Sj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GG6NUBh1vp9zmP4Xv5HUkV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSu2YcDYnew86wP5yk21hn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H2pifpNH1kjVwqth64qAwH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ybx1VaXBDvQCtRPikJFEng"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8KmmnkEZagJr8j38reWKW4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PNovtQYrWjmxnpCRycVZ6E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VqJxeipodvjzLvmZ6mp6rP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wpbi5VDqBhW5NAKrp5LsDe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9pKGAwrvnRKNPLWh9c9Jpv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Ui6oZ2ZPQHHRrA4Jwuvjy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ps1odMmeMgCFsxxtiQ3qFQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SzWwsaNrgiEaofJiqHjqcJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QpuSQGqxFSmy4VAp8dNYd9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TGmw9Do8E4iSbQXZzNXFWe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-44wyUjeBWzJZ4aVAEFPbnM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y3Hh1x9KaSPkdi3wC4ZHvF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Es4jNxYHh8LeTYFsMvfapG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AqRqcsTP5Gr13MA8gQjwcz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sg8badSiJuUcxR2tncH4sV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QKNgfyCpppAmw6bDwnaXff"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APkP8itaNQrNrGmDEGL8n6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BHuw2cAfe6dZo16t9rqU7z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HvNoVWgDafhquh98wSUkvW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VsV51quz5F7tSHeKNvh5bK"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Ytfep5F65m1AN1oc5L5Vc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VL1eouG7Y3KjuCGTeR7iWn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JY2ch4yqgqM2dDNXFLbsyv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SiSE3PMYurs7BFG11bu5X9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CapxMT8ewkiRPKfFNzF9BF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ad13ZDLMP6wgthY6ZotEZp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ZueisZjapSzjnjGYcZYqK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TnHn5mD8hdxx3CFNZ7mf9C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TWanRbtV9DsNnfCdLBzEr7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9aGLtPgD2mE3ej243hLTqs"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CHr9Tntrmq7BxiJhiPPmDG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G3k486rdHHTCdhKy7F8r8q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3R5Fi3k2qXDsnoeLeQP5h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JWugBSNqbdCJthJJammSfo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSuQeY6r8eJwwfUpie7DmP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XcfQyqv59uAHpZkpQpQ7yf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P3FWRpaEFiWDfgkcBwbHDo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KpGfLyxh5WjcnP9HqptnJM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KixwnvnMhdaL4nVfErMKB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XBejduGn8B9Kkr6DoiD646"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LPZkT5XM9Dx598daJPAJ4x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TvRPWxyEZMvwVNp1ERqDuo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7WnuN4qrqQemVB8trz98G6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MERHgugjLBTz6KuDMS9KJW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XBorB4GLCZPmaeTEHWGhss"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WkmpHNFCuZz8eBCg2vfQxP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6NMyn66VsRWdQifS1gkQT1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QPK1nT3BNGVnQR27kWG24j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V7DPifUPfH71C4U14Ht9eJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J4atFc34Ea4nxYc4Jg81X3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CPexkWJyE3JjVcdSuSTjSv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JaRYUFMKr9JqeBjovmtNuJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NJukk1qdxY7Vqkrb3V4NuF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ty7Z2ecJaRP6YcZyM7bjkK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQ1hxHifTJDfYdmWgAcbuZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xu3qJwos7jMsTSk5ECefjg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7JeLdJCuBgVa6FGQ19Mhh7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B2LGd46gbyfVqKoRPAxvp2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6hNsAwqvfsuCURgksH6ViS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MTx9srEY9WawvZfZwABxbM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9BkyWddAoFXViSHdACzbdN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5o2sQULfpb7wp1ckLoUwmH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6cEsfHsDJYYupyMFtkWqRh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VMJH1ru9EMnCkh6N5UPKR6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RWBMiLVhZjZHwkpqZKQt7C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ThfXjHXLSdtkTDJC48vDWL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2xyDa1MLTSUoE2fQDbYKoT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A26u3MRMKiAELL9mZmMoD2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q7fvo63JyhFfVejJBkz12L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Heo1LhaiZXJWrVY2ZS31Dc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rrv9u6V2TMHgFRiHd51fVw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NVTt1Zv8v4CmmjR9QnAcrr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JAcHv4sDDEFcQxVPpGgi46"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Swwjg455AAdE8zATCzGa5F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TU77RAw4MvL4mSvQ4o1Wbj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8skeonHw5fs5DsJMQfX3co"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H84nk69Ys97SK6AAgehMvq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YBENmGJA2MYPMGoAecDvQi"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BcZBdYgygMNvksEuSPcPp8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EVepMkzhh6mff4P4pU6vEM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HLvQ4KzbR8SRhjnmsDm7c9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FkPtjBod1NAFccA6yeh9jC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NgUn9LivtJCXm7HMZz7cB7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Njd3z3cFdsQKyPfomGTLGR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C59kFHeUanjd8J8SeYiiyk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PaxMr9LsoJHsxYb8kk1jCt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8w7Td65wbngwhrbVeCLJ2Y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EHsAjaHW3oD6jJwLcNYy4i"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DsHXQhR5UX6s6RSGwuARAW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GBpfMj6c3cCq5wPEw4VWvK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YBuaF9PwE4HZVk2mMssEda"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-77XMWtKxnqYXi1JpbnTgkS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F3TanQYsj4zSP3J4gR1LCs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U5RQE51GuGz4aRMPopXens"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NSJUfXKcmrkRoQV8khrCy9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ExqcU3Bp48EKYGuiHE568e"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FTYACDCgMLDjGDWriG3dWP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CGemMGy7xEgbt5HVHdBuzs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SccdFiMaHap5RNY7sS1ZGX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WQdPVcZ1PqWX6rC6uM78jK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BZNVJsCDySeDwtuwMzAPbU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5dp3373dB1ViAdZghDb89Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-39hUxhRPPtA7mU6upRcyfa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LdBsXhydHSTgcSroEKS3bk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DkW5q8L6FNoKQiJ3J7XWeQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6mSp8fmvZPpzR7uxjtFEJn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HY6HJjwQYXNKukjQueBVpT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ttJhUtPhAUyK2JvvxuVuK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HTFo3YqhUQZjFubFty9JcQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XNgxUu7DjkvNYf8ra4j4pr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DRUbcr4UUaRP9FkSiSgjdD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EAfGGmpM98e6mkBb8GEc23"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6c7nreQQ4VUeHKkYV7n2vT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QXwwathsa7HpnDfRkofX6L"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ETmpZr5snu9xyzMivmr5r9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Qhv5vrSGC5p31x9KnneJW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y5fwWbjd3RY9vAX7spqyfx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RRQ5pQbHDWSHiXFaKo4XgD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AooeWC13HhZgTwHFapk578"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AYrRmt6ftxBYNVkwS9Pnuc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sq8FW4yFkTGzcSpfSwPA3i"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gg5B6kK6URrCTB75E15Y1d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BuYFyU3UvdpD1L7YMUGYu8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XLe6T3MFRiWCuKkwW8uKtj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-25NSqdvadtiBAZGXF8K51x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XvuRtkr4uC7fXwtPzjgrDm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VuZDqPTG2vchkaLxbTQ4ze"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HSPR4htpmqRywY3iQm7KmF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WQn95uJy9gj5DMraY1yNdv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBuvaC9SwAQx6CHgxexP5q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RLytWtpE9ETVDAXKusKnUx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lp8ejF5FCi5E9TwZFd6LF8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RHhZwBK6aQ7Vz7vi8pmkts"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Htr9nRTZQdL3VZsnoBfYB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mc3uJiQkcgEHL6eSfjyarc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WvxMmZFQ5k1oJWFFZQHVN6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HaNSFFeFsMW4r1y96MEwWM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qd1UxMsUZgZCRRoLhT2Paf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RcNyPUpZmxQepXmzuSUw7G"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T1bdXSjGGD9GD3S2NgioTD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1fVNLtyYyoZJsDCRZpjNiZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F7ysZhXTpjKTTvJeYR3c6D"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V7LpPApT2PPtqfs4Zbqodr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A4eznsJWFEv1d9yFrDGT8b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2PVbjx1DoXPm4xQ69GbUah"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JZNrtqQrZdxhr6ghUUvBtC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AugmG4y6hSFYDHzCeEUGas"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5FzeeHGtYCDAiYb16cjgMB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cr67FutQHJLXugLpouCpRC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GdKgNzGnUNPKYkBdpVPJru"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LGcHFBKE7srxKTbAv4PjVS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSCvRj4vvcU9qoQTigh5ec"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5mRjUTrNhCiy8R1JfyabnB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1YhUuTEVBt1GSLMzVvNJM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RP9VScQxe9kPse14bDNw8C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HWgadwam94fzhHujkWuzng"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DGfrA4kHfgsdhyzeLGEZGa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kj9r1raMgiFnqeeZCEkQNz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QRYHnTKFhdUx6WsNYtYe3S"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E5oZ8KVc7yp1dqwjZikaCN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JWycWVx61ZTHhmjbdZXLed"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8cP6vDJznNMAMgYxeh5qCc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X23SdhsQWBKQio4Ju59BSv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T1NeqE1UDMCwjMD1XgpLBf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ChUhTN9PHuwKNqDXVvRCGf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y3ZH8LVNpUSyrNhQKi3vkh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D8TwPqqfufuDdUpEwgfq6z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9WDGPNYZ2NbvFXWcGyw8hz"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-86t6eDU5c2MtEjkWjj3jpr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JRPzeRPuMrnkHbFXRKJSpE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PAcf98QwvL5VcJRsNjCBUv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BP6x2Tz7uM6Jf5KmMrwCa6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QTNUh5oBPw5HasuaDin5SW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-URAicCyRbPRg7psC3xWzCw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CZf87qjQ3bsR7mjvCtdsPs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BHxhVmtQsNZpewhvJmaVD9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GrkLL2V7gfkT8K6PFQtjf7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S47CwVG4YdRAB3v89Kx1sr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MB1mbgbvCfi1Bobo7pdjY3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-37V8rkoUaHUdGd3fkPz7fm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UcfmPDGBBZuFdFHeFMVNQ7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZX4wsbmACXHep7eGjvwCg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sm4XHjEGouuw6RrviXxL7v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HH96TJqDFWHeHjPMPjSWHb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RXAbxBFGsd8xYUJjJtKqES"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9M19bTEte9HCYRvMX9Y4Jx"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EDs1xApTtf2yfvci8S1sU9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6pxLTcgcz58XJgnCyWNgMh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YCMVX7mva3xNLJhhq3FpzF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UtRnn1gVPnuvqQgSfrhtQ6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RGSgXWHeDJtKVZcXZvhdb4"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AWtAUejHHwgxrB6rK4Hxz9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DtqmAY6y2jVbvBqXowZXjU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2JHKdy5Z4ce8AJy6FSNhGE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9JhAyuvHXxTqmeoEVbFEc9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JJUNQmJnq9FWkpXmnPmkUp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WYQvkZpudyCAqnzJR4P3ct"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CxVuftwTutdGvXXCzksDs8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9UR2Yhoc2usPYxS2w3o5bn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3h6szXFmTkghLvARDPHu7h"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RLKyU7YqrE3jPh4QZ6EQtB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSwuaccCBKGgB6BZXpvEix"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9NkPWJqDpr7WkU7WsCTssa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K6ukp1Q7Zt4hAPApQjB8FW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NvSKFBhDY3HmcE4pyJeaKE"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TJ3xC1Vj2995uj2u1uPYao"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKn89AtV1NbUDg6KqeegDy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EPg3dS28ndukZPzaf8u5Sw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HyfFyX44uDjGdLvHyapbSG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UrRM1VgQ77F7Ba3wr4A8UZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VZinfVcWkF57qszA4n87Ky"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WPv1yaARCAhMH3x8k7WYk5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JkfiosAjq1Fr1HTvD6Ab72"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SSGoGwjmkqhPvSzowNXn45"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EPmNTXQKg8juCw59274ZVL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Sd1aYNtakVXtfqDRFRw3r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32xC5EoVdYQ4SVnBTE2qYb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UwhuVim8SrPRP4vPAa39af"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HB2XE1GA9Myk4j6dkr9HnB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFUP2CR7DEvHdnR8ky3bTD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1iiBtLWK5KeNygKsaJwA5W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ZG1bz6S9hdG32vm2oMY8a"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MFB8MCbgGVem1Qaz4eyf6B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WwvoZzsuWdaNQ8btdennTV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WnVyrtRtp7NQVKVENuCp3H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-84tp1HREPa7AJWvZYr6jsh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GpFxgivKv55Mc1F2gh5BiA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C6aCEsTYjX91P2r1R4K1RU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FHzXtWv7H8mCgMuKmTduoS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HcWdk6VTjqvGPjHGcD6Toz"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9eThSkT95gcwgMervQ7kMZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1suAAVyb3w8vXbNRiiHA5k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VxfM7JT1GJkQ6zY73UDuMv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SYWC89f63DQUXhywhmrMou"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8RA7gfFvLH4J3QzDGEuQ8E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T9G3q6ZHKnfubfYcALnVZ8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9jngvceJaGFQRXLRDUhQ9j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TTR49MdUZzEVyGRniWK1oT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CnDVtHYBEkXrgPebpHxsiV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4HGMJ1mgLLW7mEHKxDZ5X7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QGUeBZfghNH13KVbbRG4kW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KCNpAhd8NBBTGBsKnYPM8q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDFbq1K74nZxwcg5ZNPSw1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-48A8S5myFWtLsT55nZcGfN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ud4gEC7orzg9keemCvt9ho"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KqfTdvuDtwk7NiTVGV3Dwj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PUPmMp1fMs6LYQ2Bvkv6dF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8w1AFPbth5WzaPEvjAJBLS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KrMtzG5wpgLPm3aCK8ppA4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9DH5ykQY1VDi3t22q8CiQs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BSq2Pw7DHgkAV6tkTHHoZU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C6F6Cgo2GQGEyursc6iUWU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9SXAR8eyDKs69a2bdq2TTS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AHYGiDdKkQ3XuGqYJp3euU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2BERe8S86Gaumj5fQLjYgB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-18fsvsucgKdhwiphrgMUuN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kfc7iMLp3TbXjNrQXbwW7W"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6oQUtChdUVdKPkM4kanw5m"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LE3HntvRuhiicJafuR9Gem"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1r23rgQFYTv41nEsKY6umo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HxgdjwaP81VLTDTKUZgtqy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jz8iTY5ESWzpjdTB56Gmfq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5hD5xTyz1hhXP2BJXFvDr6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UFJLwLZbMypK61yH2K95vT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X4Xdqtvn6BJr871SP6gZKY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AiYuirm6Zt84PUL3gFVe8Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5St68fMcNhxMW7TUVTqEXS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AS7gpPQmdXmHkbYDRKKZSP"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALjga9owyV5nEpDnd4qTmG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2HytLZVHb2nN2nwY8PGnjH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pv9KmgGtsxtwTMRnF7CKxA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9MHpjhXbxyDvpgbjMmWHho"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TH8x7y8NT6t8rVaR9zvvAD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JXNuqShfM7so229AXh35hD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8CfWFmuUbeyzFmo53Y1Kg6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2tSEJVvqt8knCVHhUu5kt8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-43QMVoiG6JycBCVBvWu6Si"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6bo8zFvP3ZemKfYQgcq4J6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QaP8yKj8Z4TACPLh56HFji"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8w7RgH4cNy3NoH4QAFTidU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KamngxHJ1oVYMEmnzpvnUG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4JhsMkkRrP1e2PmRMVWp6r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W3oAS9xrNdNfiSHxFq4dka"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H6hU5a4nQhQ8U97FNMbqTg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8smaqoBcYHvuRBa3JRMmft"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FKeCvxsVHFA8DJbdEYmfWk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TsQ9bUEYX2rq4a3orYDy2S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8MTnzWxFAVq2wKhCPHoPyh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VXaodcnyieLhxTt2xkYBu7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nv65cemrajYUKN7VVGJ4z3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-28M4nHHAwBZb2cShXjaJ7Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sii6tajwoo37fWsCGc7RAg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5GbAiJHiEnBUA63CGZ1iiV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6eD8fUh3KaQ3BEMLRKe8y6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9YkTEkGYoumKLui77de9Bj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T1mrEUdaXoTDqrmxSgtzBE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FuehdHXEcbNV5Ni5QhdG3F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BXwTGFZzA7qzwEypwT43JC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JL5tcUPWbcNezsGmyUtgGc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2PjPyB9yZsoSJbhXqVBAfp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2mB8PodQ4eUCMstpwgPWy2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fok5Ujd8bw5cNHDhivEjCr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1L1Q1YGns1qMdrqBzUj6Fb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KZrp7mZrNGBbkiuy8yY6a7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SS5FQsmZgk3mCkVeGt5svi"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4BfcNsGjCHeNeueVymWZbr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMkrRPPtSPcQ7tR8oibivr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7epz5UYPcY7roqTZNPPy2E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EAUhzVnpEeRGV5Kpvt5c9h"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9yswHkmZ9Pgbq754jmX9RD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JuWLwZH16XdjaFwBFZuYKw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UVchyBfYVzyD9NDjeTfewT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Pc9QtH6Rn1n2iNxzEvf1R"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UtRtM6Z5eu3VbNkWdZV7o1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kwrq9mSCmPFER1P5Amp6j6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3SxkMCNFq4DN1Y7JG4tz9F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XFgcNryXDGAnKWZPG8xcV9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MRuBoL1Jw2K21iiQTko19a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-561brwHFee2feGw4MxjDuB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MYBUZEQZ3FwCPJky5HJHuu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TWfJdc3Mq8xABLF4gDp9Ka"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WYLEjor4qAXbW9BvZFngvq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CjHZWoUGxywQ26Hy9AXYUb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kc94SLoS3u3Yaa8qhadYB4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BYW5LyLKSqkD91GqiXoQox"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4tnVim6dCJVNHVWNMhq1pf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GjnCZSA5TDUPNRyqjAMAcr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WW1SfZtWexedmVJjJbhmQE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CEkZFTp56WYV6SacRVJziu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-21yRJ7kA2MLnwfJrodF6s5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PyC7DtXA8qkTCE8MpjLpT8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KFip4tVXmBYDB6rFhyCkk1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BhFiaVtK6kLRLd2R66qX57"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFb3c3w1YS8ARcgzBH8FsU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFtqjrEdBKjmP7koyExUZn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XD5dACh3av58gqfYGmQzE7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Knx2fLYxtEEyEtAjj5oVmW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1dHt4pesJVCUrVijQZ9vPG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ETVvFiML9wSZebX55yfoU7"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-64x887gfKZjGi3KXdjXv9Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NRki7K8aTHtxrbMtyw2pMN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XmXknTiCZruZmHDQ6vV391"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5sMKxipwAN459hjw4L6Bgc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AP1ajfCkZ7XAV3aF2PKuu1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RaSoo4rihQbqE7FyUSqzGE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q84jnVV1bhuuhGy8wB5CBx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8bGdWaxHyL9JffLDD6zHfK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3wjkzznh7hGLQ28kU11pV9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SSiciCTCNDwz2KffnUoDMw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PhgSgDWNqTJUMRTzuv76Bb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7PP3j5mzYhKn7kz4YoUjZt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JV6VxRdF693kWTZWsUvNpk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JHCrVv3Pn9b8f4ADGCtVw8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ld6fZ4uKP3uzUWRz5uc9WH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5H5k88SwRLsXceRk1CPQUq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PpM8Ahp5q2znCcdn1TdXro"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S6MdiNiCuVPLic7N2mf4UM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CrVBmMrWMgrB8TZjMh1RsE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BTwEBE2NfmPChJt3kuGD6A"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4hqh1sWqS1Lk3xuzd7bYPf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1KLhMcTtn3CvV8jcAwbvDx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1QQbAojqRcBw8bFAZ14Aik"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3feZaLmzVUmT2xPYaWdqaN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7cxusxd41NnZYqPPn3k8qS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KDGMHTEizXJbConEiWZfZy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-16zRciEmnZKJvmkUDS1YqR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G42ZNDrGAFPkhNmh6CLnSk"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WDoZUxHU3HhDssR2VZYqp8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ChgVRA7o5CuEohtGWHCCYC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hfj1RosffYjaupRE4FMwJ4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NGNwWp9HhBNJMzUgdV2P3X"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lk5XZ46EGyZcKfRB4DqsZi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PwmdiCWdzae97TFfukVFMw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AmfTw11EE2mPoqFs7o32zv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6BYMxVqAxPAb1FkFLTfyk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XXDUcz56dn3T6QkSNTGY2a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9UV8pEUtugTScZuGRXyZxJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MumnNWhjeCyWzyJi4Xaovr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JsUfU4XbayvoCBzc2fJRg2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YNWV5jTQRuyD1LfrPPcFeS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GAexx9159NnPQuoJ3SV6nN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FdVDjDJgMN7vRzFcuMT5c2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9gcL6it26A2L7v8KGSd869"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VuBHrnPWjsZ1Tk3Wfz81kk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HJAYi33kCNh5T2BwddMvvQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UUhj9j6peUDpb9as5W5E3r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-24FLjUThbMJ7YvVVHg3hnn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BGVu2xikT29NYfpeUEXbHE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EKbLgPnLHCufp2hdTQXMhK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Md7NveezCboZGkqV6miQH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PS1QcCPmw3QhT7CJauRMFw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YC317whpzBci9LxgYm3pU1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VMi4k2Ze7u4ZEoiTCfGbDC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4KpuK9fqdSWbdNnrQRqben"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Do6tMJMV3UvaxDcTMeLQoF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bu912hK6NgZjSGJM5upRuQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M99drMXpZeDFtKxxCNh5mc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vx57REAV3FYmnfsa1uYmY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NR9RpxUGuvMVfVqRBEtEZe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RTAUShDzVeCZz6Rnz4vpyZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5cfG71s6YQGDCHJkHqqQAW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6pTDFSU56TkYTnycRqts2X"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EQXoZFCEBe6gEUxb8Lpa9b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ZzdKU7aVyYJrHDKarLx4t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J12gcxRidJq7bv7Qe29wJd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uh2n3CdyvctsUVVSvPFxUt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-67Cgh8dXQUVyhiWz72WNV5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K3mPp11WPuuMmSEfoqbJL1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VNtYkPwerAfkfPsWS94pqP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B33xGUZJPok1JPAULvxsAR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7jEzeVtcEyJdtmCoNVz6JM"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HMdLJawM1aTnmq8XvD8g7s"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6eR4uj327AmPSmwNt2VCW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZPmx538zLovP89LpiyLrz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1FawnFCYx8CgXmg7sBWEWF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N6d9j2WPvTKhU8rhbpjkdH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SQTjy3CTnpoVA1zy4hPWRC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VwPhdEacqERgcK1p7NYF2J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZcbMB5yAXeUU5fsd4xD8V"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QabwK7xBgQ6bhzWgzqyQHr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UycgSY22risqEkd2SQ4iVd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CmHcfjZrYibxUS5T7Gc4CS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GAe239dcTiHGC6Y8qoXZ2h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HkU61oiH6rmbDowUJ38e45"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RGegyMc9NDASAt1SaxqtZZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GfDVduaXbAofbXp5W45FUj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APzLbwKuAyaBP76Zpq8Eah"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Hzc47hXuhA5rbo9sV4skN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DNHLNYFKguuJHgmiyKiUku"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ANmxcehVbmTd4tEPHx8XBc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Ju9KHkemWAURGZxVtBYHn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QmuApa6Aw68UZGKP2cF6Zv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HnZ8AME5uv5vTNSV7obm5y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ECm3uQHfSncjLCa1ctqq6E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9n5pLh12hxcKHNNDdXLLae"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2Hf3vVJPhUPAFpNbpT72a8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lj8jwTgPF73VPV8qVhsFKL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JpuWdZF1EHKhuZSdEbpZ37"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VSmVd2cq7bt9NMXJ8CcpmJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KLDX2xs6Ye2S7h4vBa8LAx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B4UV155yMz5DAjMXmkvxTe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SfGvKo1zNUv8tBu6aBVVyE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MxR3LgytWfuMsVDbvAfrZw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FbKV8qnRy8UzU4jzEN1Krv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FSKdfy2QKP6RJUCmJC2zWi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7paGTUyVBP7kAEVX3RvQWW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HSDBrXdpCz1ycgxj6uSNao"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6KUhC7Pf16HFufZ9xuHjnr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JcCjLpZ5EEpRoXVF1FWUgY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CaKuyQ2ULc7Qdv8GjXbmNy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ziywYyCCXqLrKkFMmkLjQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H34NydBbNKhBkgVnXLFcon"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7vAAiFijKzdVncQUyPYGBo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1fZ5EYyBxik4mAYtZiJsMm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dg6SqoFEdZfpBcw8jW5M7X"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MuFFSxJpiVRMoxYKiU2cka"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kc5R4xTGzJqLzm7W32AL4h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1CvJsqZ7Tqx7FpJs726Bd7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HLw2o1FJ4tQxrbJec3rRUN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JY3BjoQTJnsSkAUpRNMeGz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FUZmKtoGjYFwyuegZXG9d6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JyMszx9ZYsrsqSR3b1K55E"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9g2KCzd3oVzgUDZiPvAKR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mu9uajvCMWvshQ6VxZEb2s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32WzCeG7S3XRaAW2dXQ9BX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ybTfLtxP2CABTgvofH6PC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBWrZjbcpVk6woEUfSKCM4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cit7RDDJsAK9AJEpDC2BmE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DxreQLaWmszXpdibmyJNXP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PbrVZr5gdaGyPRG8eMkWmW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AZD7FKTApv33H3dafZ9iFR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QZAPmznLQeAsfefpmTBmni"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KYxabdwryzbwtpc2sg9jUV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8vSh89am8N6PXQvH6w8A3s"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3D4kDTW1n448WT9mPPGNFh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JVUAtrAH7SAnKNoh9jNHic"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TyiAiSWuKhZNQMZ9PJ9CUN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Qn3Fhk9kRRDrAxzsvukv1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CZEYynaYZrgzHMrjmVKdQf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4wnb52Bb9cx9eMxXAugbpE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q3jgoHe8ZoU8D8bxmYwaDv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PhQvsd4twWRLHX8GjMk77p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8x3zHicvrA3nABpYNNjsSK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XgU8xV4PaqAgX95kyiDJ1t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9qjaVn61dgSjpMidhpuHhy"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LpWpQgqa55R883xEWBWVhY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C2gBDzN1wqG5FYCEDwUHDH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MwjwatTjA8gjgbiLK4REL9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AiBy1iJPcaEkft88YcunBu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QV6YBpW5ijfFfobdZJiHCv"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K4UvaLZxKJDJ2xorCv1NtE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U14PSpNG8UscZ4XFuqiK8r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9JsxWUd6i6pXMpZ6s8iwDT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PmMbUNQuw6GyGZ46JRGcoR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1r6iouCJVcpnfK9g9G9Wd5"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UYJCpGhqyFLhgGbpAgs74s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UWLuEsjCfSi8ifVEUnRhjM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A1izyiFt7Jyu4j9PymBFR8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1CRG5cEwTpX8GGhTLKvJJa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VsA9BXgZRsptMF4jakEUcN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H7jh6kxejr6wcVN5JDJRi7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3FdWtVT8yKcpsgKeXLH2LB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JMr3fU8kfcXYU4h16CmrC5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5VNJpC1mug5HwKzKJGSeBd"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TcRY18LDvS35FRdrBaZ14r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1TLLkU1EdmTQKp54jwDGgV"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NrFX5QwPbLYbQh4UkbqwBj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FeKyyEwXnSKkjSnyuuaYgS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FdPY6eJzbM88WE7ihzLKDj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jq4veciP2xXaCGFcj1yBhc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3eCpJfSWhxXexvTwk6LYNt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TScpUJ1gyPrua2zFCVouEt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PyPPoxNu6kVJh7aLyykp2H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wrfUTD7uYZh8HeSiRm71v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sf92SjFN185wxBhFoNpMTQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7YR9AFQyGiTGdNtRJYxMgU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7AKsKQLFMfwvEtQKEwt6QC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HS3yFdKYvXWKgk3pen1qbx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wp5acc6721DiVLhUqnUiSB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S1fwhDN5nPVPQu3DWSHo71"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DDvnWVFZJF21Xvj6ZDYFNa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFm9rkp94CUJzBV4Gc5PB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B9Z1XJdowvz2ZbbN5QSCSF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8JYUPYobEwKcm3Q6KKkB3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sb69Q9b3ZeV1uMBH7VjxzV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hgn2XgE78uPC5tbJzBvWjN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7W7GE2FTAzzT1b3t8C4K6i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9QggRouXURR3N3S1MwCBTK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LBhSz1fv7itdePuo8G9rsU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XiRHcrEuWX7x2kBgFMYisu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TfAUrVP9JYaidufDBp5e4w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UHvouQmCYXGt4vS8GQnvEh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SaFYtTQbLAVL7Dx3B5qS7x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K9oAj1crnhgaLHU482H2DN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SUAfHnorrvj8Jx5CkD3uTe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XC4TMmALK33ZgTZtk32VaL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E6eK7iMdA1MidQCCe3DmyX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2QHwwYmwYjhBXmqrDhxLDn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D2S8S3kLH42WYHMuqrzJwB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WJjXignokmVGeuhoyyMFBw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ULmaGWgQBjgjARAfBxvJEq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4pBr3Wktgjr5s5b7x7hbUc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VfKzxfT24isH3U6ABTM1gL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WG3YhWM8SwUgNMneyF4oeA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BSFfSVezmSQXGezR3JX8ix"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WEWcryp9P92V6WMZP56vC6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KFk8HLVyxa7D7AmRwB1Fdv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5HdeXiRxXgscsMFaF47GzN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APVs9NK9ZsxLNvXPFrpCbM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4p4VYWhjC3mwgjNfWEeW9M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-65JZ72PKXPvxWR3p1d5wTE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HwRFQ3oBkL7it97zALcifi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SfzJWfAjtafWJnh34y2Jod"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uvd5YqXwQuvYmAHQXDWNjb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C3LCwArNBpN1h9k5FoHJEC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFdLAv2WpSMTK56s6e2Nxf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5W1vjqycgoWA3igkXGHn5Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PhYxQxUQTaB5mDKr5WJAN4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3fFvLLgah8Q8SMJc8Km8TD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W6Uu7s4Mia5kyT3bQQvsgT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BwNJ8SHmV4sx1SDBHF5oho"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PLkCNhXPLtN7knP1PPbwk2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H6vt8n3TfuZZvh3RXPjKkK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RTJSGP1qqSV3NdRCngFK64"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FwbgHAcnaDo8uC8Qei7Las"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K3z9VbHpiAEF6THAHvynhC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3BQWtDR46joLnydEyjNS4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9xEJoQsR1ZGy3rzE2QKfWD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8UpL7SVCHqf4VWRysSHWnX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GpX6QXwgSF61hbnFaEudue"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9G7mzam2sv9dXDoBxPPYL1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5rkJevgmKCnKtkoDfUiopE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-81k2bmXJXeq9bWHY6TxN7F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SN2tN3Mi6JBnrCpqLtVico"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MSc4QCMQG3641TnaovHRim"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9USaYBLkkqm8e1V1phBSW7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QJ9A4S7HmcCYwFYZwEdtr4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BPBBPLx2pdMEVcNL7EeNcY"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8QQpZ6qmnbY8g3YfcBpoey"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2CcYby2MFQTq8wX3BFGJr4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V1zCNpp2KGHH3YycqQu3z6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WkEfNM2uBeDYqLkJqV2fBm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ahmu4BFFHtT9HvxavwqAu3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ao2YeWhDmKGTv8bfLM5PLe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WxQe5z4Y4m6e1u4JTCXvip"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4af4P9uxSQmZ5neFG8MxBM"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJmy3v28TxU3JMZ78u4Qb4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-891pkuC2FTV1yozK7GihKi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LqsK2DnUkCJfWnkjqT2Afx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Piwbvda7iGNZmrJ1Un7n39"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TqXPTRbPCYq5CZTn3c9Vqu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ACtuWm4EPnVnGpHmvPEk7a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2b71BkcagUTintYEBLHRyX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QeV2Ao3Zx4Erpzqf4a8QGd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-15SivkS7aM9v414bARMe2S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-72yoxUpAJDswJuPPTLrBve"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HjqRGJnpZwVKHfVvQ5iCXe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BsSwhnQqkb3dNpyWyNjdsd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Aoa3tZEDAUv6Es3F1NoyS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JN3yvwSZMhZFAfChjLiPzf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CKTCnDjvXR72WHhDfrbkKa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WVzMv7VKfwKmCquuhYEoHf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CEmpgpj5yoop9sT8pRgZfc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mj8qHfzqCdB2wwZQX4Nbii"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9GztWq6eiTa1DgYRfYwDJF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LE6jD2wmJMp3XwLxVYjX4U"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X65jEHmdhSwtHguauX7dys"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ei8aTVwbi4rPqfS2dLF8Yg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RMDAbKGSHTSii99ZTfe4rw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UvcYdqFpm4qeDy9aYVBHzk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G9E9J5VShRAPnupguHRFZG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V91NDkpK7rjNHqQZgfVQah"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8KTFCTo6RNuxqNVY4wCFA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1BAKd21BTQvH8JgLv3KVK4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RTRk5GcHerXquVP3M7KCu8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NCXijfyjdwHMDmbRS3ZspN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VxoBuGpGiRKvAdX6zVC7tc"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nrc83MScCAmSwAkRWqHFhi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RuKqcBuj9tZyAQonEhnGCV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MpRVeXFFYed3xv8MhoRY8Y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F6hj5Ef9MiUtBwcxk9SyHz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T5MiK89rvfsrtXC5ZpKCWG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NSkefppaHs2fioe3i5Mnhr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VpAGkSMg4kWe88ND9dNGFf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xb4Y7LwzBXs9UePMM7aAR6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XN1aa4Hkuyx9H5jMoBrzi7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Juc98RobSQ3rRWzrJj4GQ9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7cEnYMNJHhks5pF1JKsyPV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FtpT51VNDF6rRYxheH6YmE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4BJiAgefzt9yNU2F2MXYyg"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YT6eXbab9gJrLkm6PAjTuJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-54VAFF1e9Li5aWa955EKU8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EwKiAXU2DfcaDXD8iNhyGD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ZmD5Mc7zVC6M9pPaJxgsa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VS7nPC3bxGYShcUM58gBw9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JNVhJSAk1Wtzzrpab8Hu9w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T9AbgCc1TGPmJxJAk9vWBR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ajUgLHGLa16UfXS3bmGS1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AWHn6HBgYii8rA2NfFuqXc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gg12QCJmAt3thWDup13upj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L3tAjU3vvStJBXRNxr8f1n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SN4RZ61nyekJxWujuKTvdJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EiDvQnajPn2SWCig1spLsD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VHjw2AtNgwh3g283Ns2euo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G698h2xNtdowJ24fJb9fkX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ASM6swvmcZernTfZd8Y1qx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B5pkddox2yLZCuZHLSdVT2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gi679ifijFVMtn76tZHtq9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CyMwqsm9weF3pKVY9EDbxT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ejx94BEawFSz8uCUg5VY8g"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sye6EXnb7aNTAWWThrMmnH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9p7mw4nD4H2nbjqxLSgBUN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CMp6m6v7WenGnzmn3H4a7n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N3dJeP5q4Dp2ky9yt8zE7H"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XZ8pVKU7q9MX54P19BFAdG"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XXPGB1y4gAR7PRFviQnXbU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HhiJPtANFjDAU4K1B9XvGf"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JPGRFJpvCNsy9ATEvbdTn1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CvAriKAKFy5NXr8GMM3jZM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HSKA77AwL1VxQPvYGU9Zqc"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XbxgNZVeF1zufUPCTeSB5Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HPhY5fNQgv69rjVjb9D6UV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1DEjML7PehaF68JBKmtd8P"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BEEhbRj3uuED8XP77UCQM6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1oCZw7moXGmXAtC2idsULD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T2W4oaQLbodgLntw82YF7R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WzasDAa4CkdQahNzF9Bcyk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-21owbBUCawXArKqdrj51dw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y5FGsBdZWZjP8tF7H8Dyrx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PCCrfnJ5GKoVewM22tHues"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kw4oBgY4x8mMyxEFTG4g5t"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PCGo3reyuAuLaEEmqDGRy3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YBs3FrwAGfjxG3MdtGgygD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-91AFgJkkZxaruuiAA2y54k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nsg7Q28sQB9iMvcNvvv3qf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EvByUeSeSmX8Ex5zi76LWE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RekpsBQwqJP4T1h39PwVgh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9FBLd8ysQCeLgsxj4g9U3o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KZmtK3cZiwyiaFYuEpdHEg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MJFR4Vtig5RVf994UFKvru"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NPPu8fxUarvkTtudZ8HDhS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NCnTvtkmuwxQ3UFcorYsyq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BnKkgfkTzob8cfbFyQe4Do"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QLwKAbqSgjgu5dCwuCp51g"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3hcKvrVXz5Ty5eChn19WeD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UmiPZBtyJqrPqxdf8hfyqu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QXAMAK5Md2xE4fUeKuFs1x"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BLHpJFVaYvm5SnDKE7Zk6t"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Drbx1bytdWwRuRVWzp1ftR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NWCTSjH8FfLinz5NAaepr1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WCXqrgGwBfpzEbpRFzFNcQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T6AQty5VUhwEcVqHkfhZxv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QwUBbNQ3WBWTncfHNWckLi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RcTdFTCvioLEND18HswBfn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1YHyqTrUQvhdSSLFm3r9Zk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YF5ZTUZJwygXwG8aCG5jFt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8tzkNypvAfnjhdjPrc5nJa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KRs3osM1xxpFeWdySQJL7Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DuFWYb5ni3SWFBSco6hEG8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xyzap96o8TWwCa8TLMPobi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LTYAJz91rSbvP1HZMhUG8f"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YLDSqT3PTMLrfCv6XJvPLF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SniAcJYawsmd85LURqjBdY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6HRXUqZi3s1fN5rF79wpvs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JRHSV7KM2ZT6fqGVxDbbhi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3v39x5mDf1BsDkUqRnPgBE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JvWfHrou6WmoVyPemmknUQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFZtEPbukzhiKPc78gF2W8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mn9eF3tRiWa6xVnz3xeXiS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-99RjC9LCCgFq8fTXTSGaoR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ArWZis2sevcpzghKZPFh5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-En9TKfUxB2eeQqkMqu293t"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XYZ9xKEmdvzwxneKPYeEnz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NN5nQv6bzEFN43HNAU9CVa"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HDtg2P7jfR3Ptz9d8rjKCz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGiS38GSFX1Kxeo4YeAJhD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DhY4yvn5YqaqGFpAFMZwwC"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Na5hE7xSo24BUZg61YBViA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9ZNWN5wSuuuTAUa13KMoyv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UDiFQLUacX5vEwhvpMcSac"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CJaea6Ch6rHzSzFdL92CrV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QogR7W27wRpGJb7EMBNpzw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8n6r8iNbb1HaX5EE7YEtma"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NHFfNtQrAYzDUkNFHfPd6r"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y3VwwPHyYHFR1TMvtGrE7N"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GibBKd3YHfLAnE42J2TQ7b"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TPpnjQs8oxa6SP7zU43d9Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-STAnvD2QJ5gzj7rTD7ppWw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GRaeEaerZ5XGBu7Kk1mMz2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wo6b3C9rwLSTkE7UoZAvk6"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RNpGk8JuLzXznLjG7RgKtW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ENTaoXf8K9i6UGbcpBYYqa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1M3p3ih3BeFcsSmeWS66XA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UbXamZXo66ztAk5ooELNjV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VkP1dW6ssVv5kFL97qWnmY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PTPF2gU3YxYYbiWEN9YeDj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7mLHc79ta3sa4GEf6Db7s7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7gJTZRquYtzojeDeU7ysRo"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YMjaDsb9HXjnPA7LbRqvVt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M4uHXwzH4KMgvkKKrnEhY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9ic7iE1qzaw7Wyuoyb5yY1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-711F71iLPeoNnVFyCHATdL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1z71ukatfN955oYFCjpKiB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6MjmGzhBRxdM8MeTmeJvVC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KDeca6z6zGGHbVDssSwDCi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MMxgEd4qrgZymtYGXGjWSv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QSZd8XjmLWTX7KAh3xqPkE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AephAYfoGjewyRmMy5gAPa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C3wEhDW1VWPwEvmAsdp1zX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7w51QFWUR81PsmMxLZJn5d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GEDJfQWa23pLQ7m8K1pSu7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AV3b4ToGbrXxeVGsEy4k4t"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ETooNXqW1FLSPCs4CGgseq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D5D4JTvnySU6Dj5sX5MML1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LtaGfgYg8bpkhPZQU5pnE3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QX4bisNiFasTFFZGyLtymf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3pEBpEEvfzG1spfRaNbPLn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ECcc1SJzDMWb56nzBSonn8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SNcAKLvzGyTG3gaXHKSya3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y568Wm6MnFzZWrLPdGFHf6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LwTLJYDWJnncYetZ4CAmTt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZbtSrPMbFkTVf34FDLL7i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WfDPDdxDEZ1fnewaw8Rgrp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ToHVJz8xWC4vmCmUehZ4zU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-412WFdZzWurmJMSc7RbM1e"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S5ZXWvFTRcgMABNwUEdMtB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YQPAT7noxh6AhyxvbDNr9J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SBXiN6zSEEQiqc74yHjXzq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-USHP8uBzyPPpzcf6okVyj4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1avCatJYjTDNTWczQ1dY4K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K2fx5KczNjiQLoZLqQVXiA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jp4RkJog6o8ohQFjRwQ69H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NZmhxdnKMgKp8mLVmZTj7j"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LmsC386ULrHB2qQtAzbiMj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EQ3WA2evXJ8wXrqXyRYdsM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F6nucuLkuH33kerN43nDGh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wq81xnh28pdYAnqeaSaoL4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JUbSdiDdPmiQdfGsNFSpCA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4vKux3NeF7frGMwrYrjuRL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eg2EzwJJHg8CoV1nZ7HtCE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dt1tt16TWg8WxVm3nZ2wFF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6yiMgjbs6C21KqeffswaH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PcFxRxCQbXbpGvv8ssSvem"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VSKAsj8mdmoFJwnVpDYs8T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5sAaqMw7DwweNrASs56FJv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7gKsYBYmUQjA2XbjLCQK2y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-53Rji2w4rQ4UhgAR7nma8f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M9vPvmiXUcSFHbHg2JXuc7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SUnegDbAoKESqQ5qqkk1vc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7DunfmqPy4H3WhsJQXDjTW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LywMucW7ZzjqM7DRx3qLNQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UdHTPqS6qcLmVUh6jaDcug"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6idCaVXFMcgsrv6dRzRet"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GGDTW7NMhQS3sdiNz88HZ7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AJsTVyzqWPp2JyaZbfUiYx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NQgM4xiLfsv8jaot2qG3d7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Co2HSRikDxzJiJVCBGczFZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WHYfmAMFDrTKKEd1pDqBMa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C2R4dnwKZ7vFquBa6HT8WN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGRKqnjeGAXv463DGST5uB"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9dDEtSdmNtMERqHQ49Qf5n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Up5QMQV4C7gds5Ecsyodp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U2ZbuPKpRaXGo7we7utLPJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-78nyRxtQA3ZxhPaFmRpzgh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-29zDxw1va1dPRLKF4uRZCc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EAa42VmHsE8RH3eYnnvGKo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6v8U3qRLLfgBdCJ4Q8KrUT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CfHcH8mJWVWvbcUzJ27GXH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Err71KzaywjpT76tZHhUQT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LTsvzi9mZDE1yhCrTEu2uF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFVncMcp27Ls5uuG8QPUjR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wYu5DbSMnxYedgY5EaNuk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VUioVCNUCT5HwEoWQqf82C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J9JBqwPK23kQokBd6hYRnu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5xUoBa9cnSnY1QYY2YCGqv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HZQNFuQZ5cuvBk6yUBsgZZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B5BBpWbJbMNXrfhu5QKHGc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C1S1fgnFcPMu1VUeZhTgNU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UFCwErHHi77v4inNegLSp2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6KUj148z6RtrYqYT9vRRMY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YCPyLixCwYfJC9reWzfBL8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SAJKrYttP4oZievV5aXtx9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lr1NvZdfZ7iq1r2GJUkzTN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ADY1yrMXeaU5xxw72FdeRF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4tGi3iAujo71SrbutxoHoB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QnfyS8H1d7fUJW8wyMFoko"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FciCWzuCasVM22eMVBYTLd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1gf2MkLkh3KLfYDGdemWjh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5CZJ3pVdvSJwNMusrsV3QM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N3TSA3JkwixPEP9o5gp3qw"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBU5YWBfkJy8VM88rLHdnX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8AHqUW9KZdN3SRbsA17ExV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HMKhRGX5BxzTtgKSnrPYfZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ZbMTDLNq5Svbj9p8BGQFr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F5YfAbjKHpvDQ3BqccjUTN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DS8nt9tAm1o5SPuMSRzrjm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FeR1ZRxxEL6LGYkij42NqE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMGJisvWucnGb9KrHNCmPo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M2g13KxmZVkuD8XKStZb7A"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YDUWoZRFWQj1zGbKFru1tw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UF6FWKmFVWSepnCzehjS22"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UBHgq8iGXKZdNaBem4XC52"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ZYQqs4NStNZ5xyEsrTF4y"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G4VuKQqV1AWHUnVhuFZQU7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DGc9rQU8UGPnq4eXsP3U8L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RGsNr4qvLyRDacWbguBUAA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WFNcZqMdF6m8nzumEHbHhL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PKpB64BGqcUXtuaFKsRRzA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FV8Jbrw2HBN8HXsXxsaJXJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4E78oTo4i8okaTwZyX27em"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4hfAAJAfYd2FhSA6yi8kFr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3QQCcReSajywA8A8CDyfJy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HWt8UPHUvhR3SqWUze4BGA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3C7ktQ4EqiT1Z5hJE8ZDdN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NYXmqECf1tyVBEAi6WYyvv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YLzMfWXo41V83i987A1Xbw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1Lq6jYgTP36L26CThMoyD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AGNUdDwTAN6Zq9CVA47SQZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BCwvAjq8v5GSYGxALrNi3z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M5sMtKPQrQUSGJjQN5yn4v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q5U1Wm6s1FbnwwCmdE8J7B"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1UDXhApjoFPYL4VafHWsS7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EK32MHQSAQPR8nKgQuc8Me"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YMe61Z3fbAc3SQ6WXLX8Ye"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vi9sgVeYH7JkvBXEb45e8B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KAkP7bNKq8XZiiyXVT7uJT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UfhoLhRnoCqVKZSrs4vHAL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3iNPJU6igdTxagRV4kcvkZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5V6hnPCQGMeLq2zJYXSCER"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PKNTmP1gnn75chDPAcdmW7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FQfyHwVWot6YGSH8MkJQjW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ss1xcsSYCVxvfzqwWU1q5o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Zj4nsrgfbZckFdP7KjxUM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SibYFmBRmQLXZzeVtk7oby"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S2N78B6LBjq9crWLbtPSnU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NNZzzDbpftmBUkM6bDsRNP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7k1cWWd51vpfgNcfK8dQ9q"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SC8GXvc5JtaEGxzTw2eQbL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBJm6m5QZA7o6BdZzERS4k"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PjNhkgnKsZ21Pab9BVZDYw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-61G2LXxt6AoQhP5Gzgyvat"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EH31yL67LZPjvnpRVjKJvj"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RxqsX4SJmEEvVDo4hd8HwU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32RQkkfvpGHzpKRq3BbLnh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PaFKgR3ZFhuksU7MG3H4M6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4B8oYUKKxVzuB12f6WcYFa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BuXjsrrWwvXUXCVpQfMWXT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C5z7LXuPPKVPHiXCCgztWE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2dD46f7aoXK4oh4ndFtiR9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7vCwLrFTX9mPnqx4Fc86jQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-67E8A99zBjz4CQA5SdSXww"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LugcUs87wuaMUQFpo1fBfu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CJ4ajpGFWyjYuR8sR2FuEq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TaNAeuarHaEzNrMvkAWW53"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UcroUQrCPLCci2QfRHZh61"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PXTDSvKq43Jg4h3MkGbfUW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LAnitWcSGkoqoKVyMeYCGG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xf2Q11WwtwryjRiaF2DQeP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PSh7PekYns8N9Zwj66F4Lm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ndFQWNCvfPeFxNMKsUzfr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KeTf5m1A6DnPWX42VQqYLN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8S1Kpe6KcRMDPj3YBZ49vt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4K87nwcDntgXcJb4C5JrQa"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DFh7spgBNtHQa2WyRmqRms"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2ugDmemUoVd6NeXsbJjFDC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4iAFngtWBewgUfEVC44rEf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ArbpGs73b3V47SBtefqcWB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JPD7XN3xjhNUfa4f4zTQJV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UX5eYkuohQ3CvFGAtDvq3Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6LXLFFxJhkZcvwU7Xdts5Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3GNymqZnP5SkXtZzRYpoFq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SQaJTiHfcdhszinovD44wb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NxhcHyJfsW33eL1szSiuuH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Hzbsc2kLHQnGXfSHnH75Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lzbg1k12CeQVLyY65iMSsS"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B1UvrhDbWts3roqNMWnpeW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EbmC4rQDCXYckbJqLP1Xk6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-596m21kovwKWQrMkg8VSy2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MqgiXcMcDuDjssfUJvExqr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4avKuZjLjhR9Semn9FQzbg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fj6hbiC8x6w8Yo8fesfkaC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G4s6VerUW3aiARUCYz23rJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BbcQECh663EgLTTeQBigEX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EVGAFD55YTigt6o33kU6qk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E5UDvaNhuZpQpLjfmFndvZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WSNJNF6YV7aeWsYxRF4P9w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MTyedXKJor7Mg51HuGpVo8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ES4Fez4qRsoK8yKFjzWtLr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-21ZpViFJzuMvFcgJNPVGrH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y3zT9R3ugFHzSt79mPnhRu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N75hPnRE7PhVJ3rD4W3w6v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bw5cdBCjA8UjvEsMp1jDtD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D3UxnhKofjfiDBLRvECmUm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9riniP2s5ufpKvgwbUzGVX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DfcsTaiYxB5kLSnKM2ibbK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3RTvF3MWdFspmWKRNGCWqb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AYi9SbjB6hK4s5Dqx5HCJA"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UMvi2g7bePgs113xwSo3UF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bqb6GHeAjoArB2Kz5n9QcK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3tf3FvVnRxjitUTSyiXhn9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CyFJxn9UJurgUFz8NrLJbR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PsMNY1KJkQM1EsKwJKgtY4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UEWKPF8zmffi9fyCZKJ4Fd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C3qb6YacH1uJ53WcvfGExo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AH3LRmQRQWvmitr4qxFgKp"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NdLo8wJM6HJx9CHdz4tc5V"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R7tzNDTpHzETy7AyNRwyxy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8YUL664UUgYKJPzxp9TfFG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4wEKJQsAvdVk8YFAkcy7E4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GQ5bmbPVZXpkyJYF8Ep9sq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A15FUvFzTe4uCy3eg1gKzQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HQHUBeZd1rkqdF8As7p8HL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EieQek7cCjAe72HwH4L6mp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qg2kNz3erirLEu53qny2Gx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7VCWXkpAiZ59RgExwAhkJ5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C77iuaYeKgzcLdUBvxk9Dn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5mqpZaQWmejXetuSKjhHdA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMCJDrRhGsWFeAkJhg4ZFv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CgfUM8csYQMV12Wf5sKQ6H"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dd9SoYFVEPoqbqvcSFojBa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WRfDAdm3BDDzons4Zr3UTy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FRuSH5hx15sj4vm8si6Qmm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-38PSpP8JpWYp53S4CHDE8h"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XFwwVJczUwLMs93PeWvFTs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XYJLtpjjFhtK6gdahVm9hw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JBEE2LfFvb1Dh3XmGb6Cro"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QH7GARAobVych5w7xXziz9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6FbGtzVQ1F8XupLBBkKmTr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fi1Tgf3LhYibTj1mvXv4rX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NbwMaWECxg44NZdGKHZ8er"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dsx55JuyREq4vsaVwj6wdq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ftaeu83dtpoH3EWQr4Kdx7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7CbDfokTm6QL8NY2Lwo6up"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W2krvh1hshjBr5FhWf28aK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HVcKa65y7W3kwHWam3fC58"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DcZ4bPugvXxRqtxg8FkmuY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMvMZcWkKiv9dSmyL58bBD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VStrqg1p2aJxusN585SKKE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R99CAQx9B9VK6JQ6dyFZ75"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TEtxRfG2otjeY9Qx8dGWYh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GejmKZ1kMxQ7unXQ8syTGj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GosqKHPdCT6QSh85RLAubk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7EUwEavugfHJNR3YRSouRq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EHcyxKDShDtDP67wr24xY2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sxc1GRUTnxjWVC1xG5CKCU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ffywebgALG2ky1aB7ExEE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LRFY8VDQ7w3xM4VQpMQVPk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WX5FeHguScGRaPYjZXk6wp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4hof3NpM7b1mJg57Na131Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GtsinQowZMyhT2h1G3Tyv2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FZRVp2RQZwYJzZyiY5dkxz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3GZvNnh8o9nPzCMVLkHPM6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9d2j1HsWcmDxiwJZpwfKsL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4PbCAHkWsibz12oyu5YLhn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJ4JvrgGog3n3bbvk5hsnb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7bYAYS3JyykVSQ79Jc62BT"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3HSDC7nZJ1JW1jqrT8z7Nw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LtpszRB9awi7GLqmAgz786"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YC8V9bgksky3fxT61nS8sC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4KTaxcSKebz6UBeRuUNDi7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JNp7xsQo2YoJJTrHrXTaFq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N3sujNmbEvAxatfMDKgPSB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VoBaDomt6kxUs9L8jCkAYc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YFi6AZriA2F24jf37UgsoF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2qSpz7ZmFXQv58NBzbgS5R"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VHtuxyguCGKDSsKvF22khz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4dfqvqKo489dZeM7rinvbk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7DowKgMQifUxmeu1VL9yJq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HC3b3najpGYQMnxdo4wme8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PkuUZGymHnxyDdQRkTaN45"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6WCuxN299su9TpKEtXy2ZG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CGfWmhEzK9cpnv5hZqUU7K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6S33CP6fCHkmL5rNCdzbrU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XaZnfHkxDsmDYVhWjBjrns"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xppmw9Abs76wXtUihKxGKx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XnSVETVrYY2YWxLRUi4duz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wc6sY3hJqNmmGC6L2EtyPz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AoQeNedr5yufZcX8Tgo7Qy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DQ2iWrMMaEToszb9LMSBWY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4D1qGFxqTc6Xio5T41Mb9Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WC3FiRC6qLN3x5chvEQWwb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UpR5wDtsTHTU4QJQU2YWV9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3sLwMZPbCrAgxsbi1XzAy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RuitqfrCSdEFE3buovtYsL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D6qNrpcoZRWKo8dkWPQnZD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AZvVTbXUjanu9E4uA4EGig"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7xpjotzNJGTzhzR9LjPbL5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6QgiRTRWuJfuLv9eZcsshQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EjGZfxDV5NkBph1Cj1bXbC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GpNUeq76gWxxaM6BN6Bjkb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DK3N5xmpa22bg4y9YGiPyE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HDUPdpHDr42GmkXgEuoT48"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AETYyW7v51Q9LJrKMitwhX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U2JQfTyvmBMQ6Y2VspjMVt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MdAQcwHVxEN4mt4itemAKu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JaMtZVe1cW7TzaQQnKpr82"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LycuRboG6kJJSp3FwD8VSg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A3fuJHBAhdxUoaCFuShbPv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3RzgbPPKRbgockHP4HAnoq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UdqaTmg4tvtEey3qiDJQcX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G6Uj54ct44gpL64w2LeMLs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KB1fipuaKzbi9DpHbn4sE6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CsAbMTgYczHzaXsWyuyRxV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7sPh9rNoLhAHP266Wk7PNw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YLRVJGTwosEGtTtJfQDuW6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-14QTnxmgChcc5VxKwXNZ2P"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9A9pYxnD2i8nE3SShruGRg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8gjEP2UohpZhWwToiBAntM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M7SzeYmU43y3aDet8mEHPD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QjpgZE8qMa3xa7ReHcAay6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y5qF8w9k8h3AiutJhaWxwb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6qijGyRTVSiNR59otpCZg7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GbjyDVGXJvKy5bzTF8Djy9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4fH6RXVhFxtgbkN4u74JLi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ljn26WfeRmtVjGtFw4jv6j"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FH7BZpQp9PhWVxThsyGczX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KQxBL184vqSUrVvZto6wdi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LebSETmnE1By8o6yweGRTv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JLbCtt2Ge5peBUbJHjisby"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3WmKuWDHbLpzYp8RiJX5Vv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ENxchZoanZUV3mWRnpbUz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TXyDmd2uTfz4kixPgSiRdW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QSg2eJzacFTB2qS63jfM82"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VXhvL9mW1Acr6t6GPZfWJ2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CoPjTX1iAVycgA9TFkKGU6"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SsnieZuV1w7pwJCdbtg77b"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VMKDPBC3NuwbR7rKfDdQ5y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CozKjihsvGGp6Eznu1pzL2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LmamGvV6ZFikYh7BBE9nip"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CqDLkWtN7T689AMpKVcGLY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BT3stMycvEGxiunMrkQsWt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UFZ5ADQcg24pCzKKU5F5TU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CQcU9HCxrED48wmeU71Vry"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vots3w2b15jzmVh21foK4q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HREp8yv5Bqsu6abdUg1F6w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XSsJN9Dd7HvMYGHEbyjNXX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7d2U4Xv4pBMezRUfwaCEBM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-16bjodBTWFXvaPFDYioxaP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VatymT664D2M3tuQtriYej"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NshzYPAre48pFNW3q5d8kW"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Av5vYXPW59vwMn9Jc4d7Cf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YZWrUp2KMZ3znomXeghPZT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rvjhy1fkTN8AZ3J8CiosHb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KRdnUdC5Rtzk4qbuWzHi5V"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S2gYx1DZshuiZeEAi5kGsU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D21qCXcey8TkWKLdek7ew9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NE6YUXbddBHSBQEr68WySC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q7gGuCaSBGc23wPzjf4Fqv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MQ2Kizxr9WFqakTqKqCvat"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QvQnkKsd9KcxmBaJ54pKt1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5h69yo51Fkw9fe7gabfqtL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7vKu97NYDixH5BRvSdbW4T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1YMAs9wT5cgn2aSc9Mnbo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FiusvXnDLKXtnvYQn1GsAj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NmEMwwfz17ea6weJU7UdzF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CRBYnCzPHHLxv9XqH5mhGi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wbevd3W9jxneHbYbQMB2ni"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WNS5mUegLHFJAKDp31yhMe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-64Mb4Fak2pEofDeNtZiTfD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NM71d1BEazbzxbfhkFvFer"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8aZvDD2s1NSXTg4mu8hrB7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3GYRi4p65g33rCoK4xLVyK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NcDktrv4bRM69kzhr5bzc1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D7DvTcwKExrPgyJjj3ckKS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-46SEFcHHwBUTx6xhUzmaZY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TmR7RSnnfChWERVs2zpYE7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M6HJDZTbEz1dTmzj1L7RbZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AiKstWgPY499aKgCY5vSUv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EcJodnvZkttoKo8fxdVaBn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SB8a6JERxAH1cMPN46AS7H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Re2yorUbmNVzTBSRGPezMw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NSsPo8Avg4iuku9WcVtL7w"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y7umHe5efeuTPLpHBDkh2d"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8gkq81zcuoxys5b48csfsz"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KtEVDKFfDNijzhNJYgBcBZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uzi6Vjcac35XdE6SaqcP6y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KQpwcb7WQqJvTxXwzGo56R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BNhQssxoia4jSBeSnvp2BC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LgXKhARcpD6NLja5TrFTCX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5EmS9ZH2sRgyWt7wkELeCC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BiHUm6WwWi6LXKiuZsd7Bx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uu1ZRZmQh959bfMCbHA1oa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PokoQZ89MFEFZR9i4L9irv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CRA4v8ijgp1VYhLDH2Gd7S"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RcVcQACHHe9QWX2omHd1SM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3HwGZYhhsqNUvMym5QRe7c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FQ4FcP8mbSGvc8qJuXDKkH"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JY1KjEyfkD7uQrYaYCkvM3"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MwZK1vWEMAitZtH8dCCwVR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6PA2LS39NYbDPiL2s6WfuE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4RYgunvwE9Y67QFFNJffgL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QqSAtnwoqLP9bxiEETsKSu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RH7YtEnrvZqLNu5aphmzjt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9aLWybff8HxLJQjRNnxsnN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HLyHcZbTNo6NavzgULvDGm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QU97x17Pww6kxVMcS5sJFh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N8VkXhPNb1Tc9ndWq7gPDu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LrEz8mgjaMpt2YEwBuupuQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5jjkJBbCTfN5oNtcvNmeQB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R3PCG3qQUtRaKBFXdia2e7"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-39LT5tAmowLiB9EjK57VvP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NcCiKXYw18zgrNGmN2PKfw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YKu3NXHrmcxGLH12yAbZCv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DpqCMFhoQmRUszNM1fUWBX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-43rdAwbrRKzUzcCsaNuW2o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QwxmWktv9cicpNHJ1rpBit"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F54saMUJgvjHsmbvAwHu49"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PnvbGrsfAznotxB79eNUuG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6TPRREdHhTJ8MvSUqSn8bF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Np7rnz4dtZQnEreeiijjb5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1QVv8sb2Q7dZmWargqEFUV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8dVvtE5zW9rb9g8mSMemDg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ah2KfrLRQrrzh2ZyvCLWsr"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8DvLuXkH1uxFy5rQ88Spbj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mr7bSwB5syc2wcStr3L9Xz"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SewFCAFW3d1mYCcE6TGsrW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PbARt6a7qDFMr6tqcfpjwj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y1TBi3AnF5FuoL7RFxkSny"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FSxyvxGHBUizsE8qbgXnMb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4hUPrUTsDMLRSemeped7Gn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3R13AAkm4YPgNTd7aL3mYn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7cd2Eu5jKu8nHUq5mG7n6p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-85KLekoKt32Co53dum7g9n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xn3xpGWt17QxneNuUxHoZb"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4VvmUTqrrTwjKb4jkn5QMR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32wEPhv6e1rAXbdt64f6sa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VhNqHhGvm8DrasLy5YNRHd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KvkAKrMw7PFXrvN2JUYYco"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3CgTfRaSzZBQ5HnA2KXhBz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xn3MiJpMhURu86CKSV9BMn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hg9wT4gQX7aCJ8KR4d7Lvm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7m9XwGvsMw5kKMugVPRQcu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6iaVqBFZamo6eHC5EXd5hL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PNv3n2dPedt2hM4VheXVGe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GrcxNffC41YLP9poG5GoK6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5phNWY7RtTV1BF1PhF9dtV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7ncLV9XpUCNRuzigNZSXb1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KF99qmRBhukwCynBtSZtRU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5f1JEMVb38FghJ8auT75JN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ntv4VUZPBKmuwkvySHQaMx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VJJzSgeC4WnhqPhL5RM18m"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UBT8xHF6YDMNX4bVHgz1ia"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GVWdVgxJ8Dxuoi2zPqaSxC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GChprmr6t9EncpJMwmWxSr"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P1UzDT6phBxY7QyJQYij8N"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LADHd4pcvXStENYZXDoyMz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3FtPXitUQuDpGtxB3ght4L"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QQHoZsn4hybsZcYUxY1eEU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBu3jW9qU1reSP29dNKjot"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8L5QtdM1fKHHnUUMLwH879"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3XQgK1dwbM6Kt1x2VyJ8E"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HKGAjkpYjay7shb5Q2N7E8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Y3U4R8Nvq8GfVq7jh8Ayc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WUH5QG6JVDaXm8nNdDcv1H"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VHAuZmmZEa5UZTk4fpsJ78"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q3197WnJHucVeBTa5WfV68"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9yzEYfHi5RiF6cGmWizKRw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HtqVractZWPap8jQ8gkgqH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vn2a66pzRfU6xBbsY22Ga3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NmBpUHbba1CEPuCfnKVrW5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AmjgZzmGa7cA6QUtcUaSMg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RECjwPFC8LPwr2weAG9Ast"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wv5D9WypThVsNKkATwa4ca"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NUD5uN3Ezt1j25BDJhAFnC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H8oSZVwqBzMyUj1GmPDGTr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XXM3x5QoZvcuAfspcSHAdY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ThzWk6FyTQmt3mZZVbwtcs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XYAtJ5mf3ZBnFTM5do6aFu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PQehh3oUxvMDGfNgfsMBBP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7YrfX492aRzkN9NCmFi9vx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ch3q3pDDSJzugrpjsFpsuk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QtqDNQzfVCQtRmtZQ4ozgc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LBndVCNKGudLbkUEMZDd9Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PFhUDnbeW32rE7FQN1mpZv"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBMewcd3NnTjBVZXKTjtVG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UoevyCPvQQoAse3dz2f91Z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9qE97LieGb7R7E6UMCk7CA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CDW14ZgnxsgrVdzJyfNjVj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PDNMDaa1Tb7HfP2NogJDub"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4V4z4jn664TzNfizdv57ZK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4byb3LDi58KZoRBdyh9fbP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HZXDR7svFtBJ82X16QZvNB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W5feBHnztXFD7299qq72Wd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VQt4WsrsgX5Lb4U5A4NEbw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cw7bYCDBLczkwFM5bt3CF1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JCU56oeusDRv8r4hoUKWRp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VyxtiY9Jyz5aNa1LiEYiaW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JRCirWCcm8KagFhDApXVj9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KWUC7Ar1y5sS8J5BRhoNFs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1yN4P7f9vB5vC3qcTZP42Q"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HNkfQ4HE48Loji6PeiA2wa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YG7C56NfC75gkb4HSgExPV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lyu7dfv4x61qKyXpvAwiup"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AtJ3cs4GbXR18vqALTc711"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7dnvkcRYu719iPWKvuUzP4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hcym9rLuuoU1PjYZykUrgV"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GZgwM7VTJtboBbAHpWrcsf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XVo4bgzvtC5LEx9K7yR6Ab"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BmdBmVRny7AnS3q34ETvwZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9D48Huj9BxgejiVTih1DAL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8kwtLrEcFkgdfcFXRt6syy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DnheVmwE8XxUYfjYueeGLh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DT7hypdbc34k77ZJ3tNNX3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GPhuBooy57TTtw8MdrQ4jB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UkEw86PpGJUntH5ue28vdx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DgXHGc7VNJXpL7ykKDZovr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NLHpDJCZzvcfu5uQnpnWUQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N7G31csuuZ4dKXGo34wFoW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DE3gJdKFtWQJhgQWL6zmFK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EWTAovoEdP2oUkYC7FV2BL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-StCCfpWFE3FjEC27ig9Hrk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EEKNQdVZvFp72WDJ9HMxa4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5UfDWJmL4PPvvAswhMDSXX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J5Mdbgnt1ign6pcxW4H5JL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4vwKko6ju4S7c29AuP4unq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SLTfPevhEeKfnsMBKBvShi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TasJJW1UY6Ps1m1DXDUKeS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P2wACsAR4qaJQKcWxnDGK3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UxrYxMswCHypFGhGHHcB95"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QkQLGMwB2ab9CEpUeu2354"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FWvU1VEA1sDGf2JGf2gtaf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YALE7Pyq2h4x2Z7qBULyBG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jhk3pJu6sZhPeQZFnBnS77"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kvio2yh9g5Xs4GqquyAAEX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LjoKGPcC4uSNmsjv6iYAvY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3QzH7Dpfn4Yn8s1eFc87Nf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VEStgbm4ymZbkz5RnR3qhi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XNrzYn8ZafAHNYqTBWkktT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EbLKDisojEv54DTCAmTg6E"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4pp63sTaSuPpcynNVZ2vTe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-87hZuwDvvFqMLmRc1TPq1X"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UvUd4nfyQQksLJFE9Murkt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YDpyEmKzjzSEZg4pSZjNgn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-STjgHQJC45bKv9AXZqLYNr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QFefVk1ri1EUCzU4mMUtTX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LZWDkFNzxZEAbSsckyQpxE"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2baXmVVnzYBQrY7z7auHeP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9knfYkmByyJdJSYRsDYvmY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NNPkThCAnsh44QtYetDLcf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A7wM8ZcVSWEydHXjgYDgUC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KPXyp8vpecNCJUuimqVvVu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KkMZCjCqZ49pCBT4yrGfpE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3fHWPP7MDSwH6xo7Yg4nkf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-19Y5zs3M1AcYcLBDs4x1vs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cy7GuHX7ddtsLAYV2iC7NE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WkWYoUxqyzfjdMhDnQxG2K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ct8Yqts1Bn1TE3w17xKbso"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PirYQUkZf7sGGX5PEY8aWN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BdFWeNxGtuv7aYVxpUCqNm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C8pommkEm642cZuQyVSw3Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SXH76tBXfWaRx5BgnyjgjG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Mi2zXsQb5x67jU3CdC769B"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JjZ8xrMa1Jqn1frooqn22t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3DJNTw9hJu62UHLmFXvYm1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VewN9MNVkjRa1qiF8zzuMv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2jtapAWSqASzbZHrTZKUak"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FVLoc4Ar3EpUF2ud4C4Pz5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WP3hjeQtFXxJz9ba1QHPiK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-522U1sJuwJZC5XJK1PJNPm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RM6aSwNTubgHcaRpTprQc"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AHwSowmRGRFYv42bPYVyte"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-58MfFuz2eVfUXbARAFLzzX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H4coMGDMuMur5G88QayUws"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBwN69mKRFku8dU7bHupj3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5a3uJSMJ5SwCdeWGYy5ubN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ESz3s9roeAoseuGmHVPF3Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KbBqnbyT9eG7xY5GRxBQZR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4WknFP2dR782BcfD2toG9e"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MZRwRGeWkKoMnrdDkrNcn8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TAQKWMXWx6Mw6zMacCVqdn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CKxuUamyLAwtzLdyxHa86f"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ra1hBQF4geddUGn9LxLjf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8HYcT2uQdNJNLGDLNN694m"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GGPm4984GVrPT7zgGG59xE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4f36NUuCcQauLBRoN7jxmw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ExtREpHgUhzU61hCF9BEw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-97HoVq7kRH2id6NHkZddp3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DWVq1gBSQrvFJhD5u7rs9W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C47g981mJbeYzeafh38Txa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DVXNV9wci2fWE1tQcVeXyP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qj8SHeZT9Qt8nrxEJsWNoA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Po3HDN63cwbTbEfEJ4d1M8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eb6gAXREQh57i96as5yyMB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9pr2wuRdDofC2trag1gZsb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WYXoCMgkCzmCmUAYW3pZJJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JiiZ8sNdzz7oV7G1oX9bhf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RpWbGs9AbfacTqJp7YAe4o"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YcBgMDwDaW6QESex2ehwrk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EjT7LZpumL6S62inBxG9jd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BNcqE4tQRp6J7Lc7fiaE4C"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y485MSgCbAZTdCjgqcdPum"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2RHmfrBYhi14Gp4o665qDa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KdGFQqHhJjLRUuVRLQPJWS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WZmYhgtEDFsmBoQK9NpXt9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ck8C5JYGoin4nEp7kZaiV1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DZNJLv7dCiU6u2KPE7UFd4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hg3UYbKm59ioahgXdNX3rs"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSxerx7kV4J11MH994Zyid"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TJkHY7sy74cn2e9dRjfRdq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WRWNgX8g2kTmbnUtdjxBfm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KuGsSSMiqa8EHpqLvJrZcY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TZNVupNKPWt2tGHNnXaB9z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TdYWYBhSrmnh17LmqgjEij"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VQ1mbqrS6mvPzEcrooyK8K"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FEHb6xYKB3NJw2ZRQ16Ls4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AhsDTKC1PvuyQdo6w2fuch"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gg856utU49GswuL7sirGBW"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BbwUp87xKqeLfXRwt2v2wG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9R76wydB5N6dvm3LiGWQb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PDE4f278kQo4LWVXVcrEvU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PcBHZvVBFV9iQeNoTJ7MTM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQgyXe4aK4GShPhmrYf2Ze"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BvokE87oEvKRdgxXBA6sPp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1LTfgtQpgk3aUM5eSfrr1d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TQDXmC4hxJxpDtoxpKGG6X"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DzWccSjnZicsLPph7f1iJr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VDCB2Ex9UPaGP7dkc2XZNj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S71D2LZbYMGiRr5HsrKLJv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sgsae19ozAKVqjiNu4fgtk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LM7zWwes8WjtatYQX7MH6q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8xd7mKn8eUDznqBL2r2kvs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8AmhDWP6vJrvEC9XcV5FhP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3t3ijx9ij79jDjs2qvUUAQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMZu84LTtfxFdQ9i92ar8K"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FaDP5TgXhGQTXk6wWjXpRZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JcoWuqCMKNkeEGSGMiAjzs"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M6gqHqXs4MbkQyn7d518hD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BQKK6QneTFXxnG1UAQM48v"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TqPHeGfqqN3Hk4GxVuAKby"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9fEXFkovKfLoVTDAiPuJPb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ta9t53ZZTe4oZviLUFmojm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NX3X6wqe9z5zjZjEiBZWxi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UxHDGQWZAUVqnid96LEecJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nu9omKa6j3GoPSHs6HyKE3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3LLETm3Sk5BGb9212iM9tf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QAiD3QzhfU37ipybfXQTud"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5U2GHZf8X4FF6KEcLg9a4A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AtxhKbPQcSbm1Dhf95dLBa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RGooaqWAZK4eZgfH6R4Vwn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PZLrXVJPzZqtL9tw149m2K"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ls53sR7BF8ZmspZD7tmLWL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JHy6c9jyMZD4fpmHxpEpAj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NSyzQQHCfxdL261qSNK6PX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F1ieXZaAuWARRZPESaRjWe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5HGASKdUd76pGrb9duqnMB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9UkhLZvYv3b4rN2M6kWLP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FtqjcM3xhAfXChmJUsnW4y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9kG1Ltk46ofJSYhZMMnqLT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VT4qvycQ9XCnuEn3uu5XXf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Syrm8dD6V8YqnvTvj2iJRP"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1XGa42fZpdwxhwj71QtFAF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XSumZsoucaAaobpdNU874U"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TBLEm89bq58sSXoU4wRpAN"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G1wTJJaUAXKuERM6btYQyE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JQrMG6w2p6d1PL54nJR6c4"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5mgChMWGVokHVjyMJA6L3N"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A7pJQd2q4HmXYcZQz6KfTL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9CR43aFH54joY6kXL6u3xV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TynrqdazYFhh9vm5NKmWCs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AnZxr9Z6aGynZ6SMAn72M6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3eRjvjyrkkVxqAV2HPzFLe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H8fR8YwR3c8n99o3J6Hpr1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RpKeqij4Uc1py4B3QGNW2W"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UnspPzSmTJPsPKNwo4EKtA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QVuFhZymDH3aHFKtgKFXtM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PsTzXqBLhbJyh2i5Nww8uR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V51ULoxVcWJ72hzEm8gAfg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B8sxMykHxQu1QdmSpFM9Qm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BLVnDXQgNLsP1iRSs4mDM6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SXcgjshbGBNc3e95iWPNyf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NCZRtvoJfoNvDy1v34BNV9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GeJsYauKzc6oK47Qvaspa4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D4oiFqDXeQQpqf5TZeWfhP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tre2Pk7ZusmTuEZDDoaj8V"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M8mHwzRAQw2YmWnZq1NHCk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3weYfQzYCWkBr11Bib2K1a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VpS5e8wgyw81w2ffaRh9ZJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6vsrv7Y1zGJbfHeTKSofLo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QU15YxoPyJjdj4NiHz4DCr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XTFxxRyr42hFm5uVJaHn72"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F6c4RvohkS6dCDvpcwHxS4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YZenjN6FkscLUUnrq6wnU3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LWoB1414Upy7rGQrDmcvDe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UfKJxcyxQvdum9SKEFXiwM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2seYRqdAaEq9mt9Ebb6oUD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eqk21bYcsk8fj2X2DdMfqP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Htwfi38qQriwawn2W4RgB8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3pXMdmGs7ifLUnCcG78CaX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RUaN9qUok2f8r2iC52zM9F"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PxXySqkrggoVQomYygkDH8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLd5u7YdSSPnUj92Uz1nqQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6ZZT9PjRxqnQmiH2cpe4kM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9DtPFCMaQ4JZBBUiZh4MRE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MxnB7A5kWWRhE6VTRyEtz8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HmhSoiJgCRps2A5Sd36Dd3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PaxAS36kEnFHTRg7bJhHnp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wx2USdc7b97e7W6peb3KZi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R9Dn6cX4S7WgYdUy1BYJ3o"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RgL7BhbMRS8nc68bv1TFem"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5WpZyczW3m6rkaVqssZc5t"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C2ZNme7v3cMBexfv4EtL2y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vu5mG2rTaT9FZN1cjiQ1mi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FBVJTqRQEbQVKjotzirVQ1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JsvXxNWyLSGfErHQxezCcu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-52FZVcdqSqj2CM18xYjAk7"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XZJPgdVgj7HidVBhZRVPmr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CTnQ8SjxmxhKLPFZnTLC88"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2KpJz1Pp8VRfHEXVqZdCC9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LiGqHjw8oFCYePk8mig9fi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CnmoNcRPbgaWx5ni1QwyDW"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BdgcGzqX4Wp3nvhwRzgB8P"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7hjqnZ8jds5fpZzdPogKxE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DSPnPEgFxtEzbYJnAcHf9g"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8HVpy32qVFdLda16MYUxs5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2CNEpKEuc9EPZ7Z2kokfBt"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ky5uK8jv9YsdThJcr5CWwL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UYBzncKQZGuwdnQr8qnGEZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5siENFNzbAjEm4C5JHRRak"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E3dLJaEVqJsk5sujmrnm37"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3sBXRndJzKWBuEnB6D8DCV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CN7MwQfKjMXwNDqZpiuJfV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TgyRkrzv6Cfm9vAJqZmYd4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LUePGiZWTPrke6NiTZFkqK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4pvKCV3vhoRDUWKiPibvbi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7HfjWyw4n8g1LEF4jg3Sxx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KTEPMLnsHn7S9N1SRv9cTA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RMAx2cdkLRvsPNtNSG7eME"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3aVJHaKoCw9g4ge6VruVBn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TuxQkPX9SPSaFe6uLgRhcD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BoE6QMuA2zXU6j5MvHntAt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HRGGdGKKCoKeXx84YnyRs7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MFWXjy2gEUfLzFw7UFBoo1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-17uwh6jfxFe4xsGt8jm7hP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PhTP7hYBo3Wak43nNLckkB"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8VqScsJfMq86zkg7Tk36Qp"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-86tzzM1hr4GJWoZE42ZPFf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lrp9fHrByP2d6NzYxe1B2z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WtFae6Ya8aidjdFCPKt3KS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DRTkoneU7EzfiXhm8AXZyY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7GDeQ3kXKF3kCbKJSb7pjx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FMGzPCpFqPzYJxZ3BceBZL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jid8azGEK8kzDEYDkSPBga"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GkcV4Tij1wmCir4cLNBYgp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALfVo4vwryhn3aZocgqryw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QHJ73KDhsjRsUrdrs97JWy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8HVed6eZ3UceJJTVvRUch"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4MBQp37FSjrzNouFUtS9d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5u41SH6r89QHRvub11f8dB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UqfuQDafzMYzjcSJedRB4N"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsWeuNd5e3FDyfC8nPpEY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ap15Bg3K3dE5iHqU2b5P5x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJGpWYgNiStcMEr6hDXqLG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F13R1pav6ZEuDJWvjziVtA"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LX1mVVtXpxEe47pQBgsSgW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DcFsSHZNJRHr7cSLCZDqDf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VQEX4yPQpXokHg6JUWMXsL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3uGKUyzDpbWDbgxsnEzxU8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7J5TtCttqRhARNpSmmBN99"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kv11wGXJB3MaNmkoCZVycE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pb5PHLNQ6AixrhuPdDopPM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AoPGHVFvLSBFj6R1cVivnK"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jv4hJmcPK5ve4sCJVvUfcd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4RBUXUjpGJ4Uw5tYr2j8uD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T7JdnUjigqUtYuztas5Cy8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JgxNXgqbs3YQ2u2eS7EvBg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P4Lqa2XeRsNcUmQDmjVady"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9X9z6ARe2YoMoFyqErYVeF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NKqikGxAAR4qprVTdZAe7i"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JS4Af69Ze3sPH1DzCMvNKz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XYLFY3sp1nSPn66UTVvdSx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RKF7D4EXLKoaFZmsyZdZy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pvs3BUW7WMCrVjEwgJCVyD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A4Dhtp4FJHSdviq2ZjfnoN"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SvRKbuLFJH8UiYmuQBTpi5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V1SnS9WY4hwPzDwdjfy8ov"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJLhH1wjfYaWHhAeAEQB8f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MfJBrBUtT5k8BqPjZt4w7T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LLyoJ1U9Fi9rsq7XxF6UpF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jcnr1yXRoemew2FuXrxBJ2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PC6LyW9ey8mjPCNmMVvzj9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MEjq7RGMqLcyEWE4HZxDt8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2YtLnUFUFCAMAd1mrDCVE4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Yc4oak7Rpy98nk59AhUeBV"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NYVeHmdFHkfTBEmJKozvMC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-61PY2f4yTV6HEEn256YUPQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-69YhD4ZLhNYZdPxuV8GMRf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y7jVsYcfWPyN2STQ3dj5yK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsW8VL47ng7BKNZwQitxYz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6Ft1zKKCstWLPtsu7285eK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3vCqVFcfSump2C45bWGwn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8dx3XojWD1YYxyjXWHjpHw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kp9xLe4sNNi6FemF6hvLoF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NmMYigW477KWhGjBV83Txy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cykrf6rDRJqAeN8HS2vGCt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NpFxTqWiMipgz4YnAwTJAa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NVopvPnJQWNmUUVeEbq6Dj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9RQP6dozFFYfJkXvcGDbL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Pu1VZkfRPX95SGcXPRN43"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BdBnuHWXsnXtpY4LHW3cwd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FSxP2LzQYJ15hBvjjSCDfZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L3wZiQ394oFkwK1LPJsbwf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8oJj5xdb9stW4EzxcfKB94"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2Z6fzqeKs8gyUgDYHVLHq9"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UWNoT86UJ1Pta9j4KnFTzg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1G8W4CViYTzXhdsCbvVWLq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2i3W2GHicK88kvHSvcoxiy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RT3B4tbMmGTMngEq8MCcMj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B9SfNXLtssYikciqiJW9jq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vdz2EhnsAygGSxq3B5Ddj2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4N5iVk7wQyTpoviAWD562r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KVC6dXLuXbPQegQnt84UgZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-77nAbejqgYJFaR4WJb6C8y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V9GKf83MySnyYmvQYwAkJD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RAWApgLgDo4TzsuhxcxU8s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-64d5ESv5kRzb9d9SsYUhf8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9hwWzTDzDts2LXvzSw5Fry"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4NYwpKfzz3xUWeof8JgPvL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JeJ1YXK6ThAvm1KTtvVqnu"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-43KhAmwRZiSiRwXaTuSMKq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bxbf9a5DE15hH41U6UJwWW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7tuHtKV4nimhfSMvRuPgeD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3tTg28c8P1NBZmFbeXNyjC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KZk7XKFfpiVToBi2w26ebb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AYkWXpyiRgC2BMKcWxWnYa"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AdVK5Hytg4WB9GNtFjabqQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RWToxhbSgP57HhUrbQFoWC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rti9uXn4U9cesgPUyfpRJn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KJqSvdbPdvuuf8wmtFazQ1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3D2RMruft9UhqYEcWQXoFh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MLfWVN98cUvBTKcDDkScnM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7W3TSrAKkP3M9BsM6N4S2k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5u4EXghQmwDFhRvvBVUYZS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GKVAHWbvPv9gp1NcSC61HD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5be3rRrJotH7JXesgudhtx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LCTZuGiwUkqZ1UMTPkaBfw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VhbcpfFGAaDkUyV5b8RXzY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VEwFYo2AhDjD8iunisxE9M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLHDCWduSGe3MzCKcbhndR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uda23zjmHpEwmPgeoN92Lb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NCy9sv24hhis4HzC3n7SEU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QXpngaidasCmFaB1aeVFpa"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GgJNSjiWqLFXCZiUnsW6E8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73vHguEhUERW3f4Vz6H8Up"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VkUoy9sz2QW9A1wPb5y1xb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QBDUywLwPeowx1Nw8wwVxy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DjtDoiZDrNqsFwrncxJxZJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CpAHqF3NJJeEkUDfqzk4NV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JqmGhDjHCGAfd31TbiPexX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TqyodeuZRt37Sk1eBjxnKj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HnTaRtXNcYHD7VAouxNNyN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HLreCbDogV8sCekeuajsQg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TrQKkLyq7fBABaSE2EqXDg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WnFbvXBTNKLoZX8zH3t1ga"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SxJJTc5kzFdv6iZ7Kw7EeB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UC2VAxWb2K3P5XC7TsMrmh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XWizRepiNQ7pwuvQishuBE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RKmTkYvorEgsKZ8LhFCdgy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LkMnzEjk2SHWhi6y1Q952F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gc8Wzn2TtfET4X8pNBsqGb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7a163dXwFhcpS4yKbeVtkK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGs7Ze3cowv6M6gANHZp5m"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RgUCJCdrmhBGJ8LW85aaJz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PY5j4YpijGqHMtZ1BkVMVF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SrrvYEuaQaHZEYA7H8xYuJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-58nLs5mEF5og8MdHiZAYfc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JdsGSoqsw49yAMn1iZwCjM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QFoVdTvuUtP1XkoAiv9XTh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2gLmHFwkLVafV4oFAkYhuW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xpsonk1YkMJwKK5o1myovc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GjKA1EZGTaexjFHVXqRs5r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KaM7nxGbNBXXTJLQ1BDXUa"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RxMn3ZXAaYcFPrEkZYWy6g"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XGQAizFaGydAG2dbf2Jukz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-23i7iSxk3VYvDpp4pQ9NeF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QUerbrYz3TAyv2jcJdNU6b"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-33tSAo8cysmXzNCQ6Tyw8t"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RHps1hcFcBoNovM915JivV"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6AH8TyBNo3Vxc9r4e5TUS5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4NGH6V8HEPsYxhUTZEkr4e"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-18qum29kkMZM1NRV2mYzqZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8oWWF7pg3VoVWxJCzXufLE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6wXu1xBaVCbJMd3x6ExJkp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QBRbYWMaQ4o2iYe8koeZ6C"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DyXR9Ag8sMTc6YvxDQigTh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1fipcyCFtvnAbmpc3dz2PL"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XsDHSPr9dLP9Euzm3zyZZh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tvq4m4zKoHznBithTsH9aM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UWL9joNhyBxSvkdz6HZH2t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-84vu6rmVWD6tTPzasrbnjE"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Cc81Gjg5gtbNRcd7LqNTcV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4BLPsbb2UjjyLecj3cMj4P"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LT7EXRQ8JZYB8ZwMAuPvgc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-64K79vjW1LCigjtG1Z5Xsz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-77z1mwxNjrnXbcwLvQsvvH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ufk8hSWDVQfaByw14qHUqT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RuoCwhwxuyCixg3gBdy9Ms"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9xPwkXEJoXiTo85bKoy1y5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MpXDuRBCoWHUeveUmpudCH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LSMoVDLxRw3Zgrw5FcdQsg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uub3BKuUicpeRX9aKRygHA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AaLKmr7544RoVPwmbnrr8v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WCHucyrQcDuKXz1nKbVpgk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EqwgDdGRZh43ibzyreqhdY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nmqyam2aqCQV395PAYgQ1N"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M5cdXb1Q3HN2qnyEm7CVmi"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4nJMBfeDBo3j61Qi8ZsnvU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-54J83eRf5rxh8LkcBHvFPR"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HykHxNGZirgK4btsQUzLo3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sd9wWFHFDnNShhoH4UVZ4V"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PNaLj9GCwepoDEf1mEbFBT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U6KQAjeiCuzAeYwPMN4aY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JBp1gdjorHgwEuExf963Mw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6VdgxELadqKz1nfpgusZWb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6wSJzNM3DHFH5aGzZsZMqF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Na5QpJZtRuQPbRtdgwYqwt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WBTHn33TobYALNTi9ZXGFW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1WqH9U3A4NKXHrLYeT8vRB"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YQck3sioHLbACnh8EDESS9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D3S9HpusUxmRKS4YrCcdHN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CFm4RFLwQBvwpFX2n25anj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jg9tMXWkeW5nj2MJNTZbyG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KaEfN2VX2RPDACbznKzLXg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MCgukXJf3tkjLCRQsPr49Q"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BgP1Dsqr2ZwZMkm1jcKAjD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BCQSsgnvEGJV1UKeBpGCRG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CtfAjMJdCV12NeVM5W5jZD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7fToLwyYYmUFJGfmNC5xeA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X7SmDkRVGW3n9q5A725LR2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wd9mBNy3s2VizrczBuR5Gi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YV6wZZzdHNtvdqfVQLGxMc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D25WGkg7DgHaZpUVcd7bVv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8umS1WRKydahDbSnoUDmDf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V7hD1qwPW2wj8SCgHzaXjK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J5aG9gcxDekiznxhytzNB5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8qgrtAoAx3qGN9guxmqv5f"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-REiC3A8Zaqcqa1ZY58P159"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6m1k7CeXBQz678SVgeuwF7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGtbPvvekCkWTZKy8fappn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U8Z2A6xw4tJRuy7cg2bnhm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G8HyWw9oNXNoRjJgQ2gKzZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q6oMfebYnpV8Bnphgq4v6w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S3t1Wjf8wdZ3nTfgEcjMe8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KPx2KxS6G7ZncDQDGruDfH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Niwb5Uj6EzUTEW6Q2gszWa"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DuEZYDPMKQaae8c8FfKpnX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3DSZCBjEzuSwDPuE7NNoNS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rzpuu4sLnFoZW86RTqozMD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4VKQji9g2CkdzwGjKHehY2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6QnpMisyfV5Wrzvqbrauct"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NaGF5LPxpHKSjiYy9ShaD1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AwL3Fcwq3KLuct7KFPShGj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M76GKDSy1mLUGbyJh1QTDc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ToWUp3borT4Ltn9mdeBk4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QDp5G7nstDcTAhRpf3VfuQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KbAa3FQnXjkPEckRXboiDe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SeYhUJtrUdoaYUt68Z5QMV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DouwG1cqpLcgoqKGGtpjrT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5rRPauUt39o5tT7JbBMcKz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tc3BWWNXec1H1mUTCWnKQM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RmSruKRpmACrUYeSkQacCh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SSXsD7bERw5upya9PhBMfe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KxbN3jGw1iQQQYUUVSPVDL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4ymn3y63KZsB8E4kuswYS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FhG3dfjLrEPCLLx3xtr4Wp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7xLRWLfQBUa1avntDZtiWc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hw3qeNJFu1oumYc7XWR3Lv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MCXqAhXpe15UjGLi9eNJcY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MwNLbvm66wm5YLTd8Erfp5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KuDAHCa121qH2UFbEj3rJe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J3eN2hcjQuRi5sqx6AYcKJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rj9zGCc8e2VS73ggpzYDs8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UHFArbiDwPmPDwKPiVQ2gs"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3bX2GqAc1FPo3ou1dFurye"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UAJu7XejCxC1bYc7wGMHbs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ry2Ji6iWy3ySqB2BBJZQFm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DoxzSFEnF3ATFHQdFJFbMC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RmjkXkwP62ryEHuNzYu2r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KT7nmPG6YJF57eaZFrRodB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1EG7VTuAfM1oeuRtybSzgx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02052_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ybbt8Maq1ThEVzg2cpcDPo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JSack2LFkFbXrf3nDnZqT1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K9BX75ZbAViwE19ScxCMi8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M5JdnU62SQg7ztYuSacY85"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BR1Er7kGm5oazigDvDaDav"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8bs1Gb1arA1WNHDd8uzYWX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y06156_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4zaGo2etSHyp86ikpM2kDX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-57U4odAFTE9SLkwdu9PAVA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G5HpkV842nJzaS1zx6VkBX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LZj492Vps5RCxJFsPQFF7y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6iNZehJ2qqjgiYCk7MJH4f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RTyScmkvTMQDgbnRBqDnJm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UPvui6UAS4KvazqSwhy4Su"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bne92eLsX2o96z6EPC1ipc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DV2oBrmL6CS1hMEqmqgRay"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RVR5GneiFVJoTXU1kUCqd9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SS51XNQkzcwDQqnCDBzWfE"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9zejfvAK33qysu9jNFZUgk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1wz1S35QcipMSgu1XrKbU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R7pjR1LZ4WzecvpLp1b3fY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TLUGGyEtT6LVq9w4oG3vFz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DD34Gwyh43qv5G7tG3AffL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dj5zaees8dJGLeHbgG1FbY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GEebcku7ry75SVJuifWSt4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4EF26uwtDmC76aYXtLo1Sr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7YRvJZBH7QQY5hoa5cCWGP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Pbz5xKG4tNw1HAsQFViQY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T9LEZkBT2v2VWzMaXsMBta"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HU5A5buJFrubtifFp78BJQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2Pt277TH69EmykicrSxFYW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YCYiU7b9YbJSSANwy1YhCj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G2hPtD7gqWhFDhxiBW6nYo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NFgB5TaKNaj5UaTDTo2kp6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BrZASpbug91exkq29AZ7QN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4qm54ejZUoMr3E4JSNGkY3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-12sgZsiLqr1Jexch1Lsgd1"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DhsvDem8oF4KiVVrf3nQU3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FrckYiLmDn8QZYioyYS8u8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LeWWStw2rM91Ko5DizcuBQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3KDwwXDWwkjrZ7RmthNoTE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SyVc8JiTeCatsPizycnrxg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BGyES2Qf59xdm7V7TeDmBX"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MsoVPawZvBZmymV7UnSxah"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vac7cserSsVQZfiBQPF71e"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Qy8LcR3gG2sTKhBhyhMboL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HjhyQ28YYaFvf76pSLb78E"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HNASEmYayZm8EAkBSsPyZx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BU3wKrt9FqWpviJ5RDUGV5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7atXrEsCXfbSrVFgkoYGDu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TYQs7ZBCe6DvHouet8JyUo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LPFmA5DygashMncZ7R8GU6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FGEgD55LUr1cvSVLNREcMz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HsYtd6KWLKvLCZebw35sef"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7DRtrDeYMnyHG1oiptpMk1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6FRV4wnCUNpEp4X9EnPj4p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KxfjkRqovUVNVxSLnC5U6a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FccwH7kStVLyKXaLb9cGzM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-87VyNXjeWpepz8Cpcd6StB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y06156_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NYi157j7XQZsJYyGyobFLm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QdsneXNN79XqeFuVRE2vwD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N1eZFnHUzwDMYFprZ2DiRU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AB1HRaa5k4f1aoWSBiPUhn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9uDuu612Svn8b7oNLbtbxd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7iLmJhjdG1rJyrGJbTBa7h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V2wTpaQaxsZ6oEEfWnHf1s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7fay1eQpwKs2N8DsmYyhet"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FRzUpVacN7JBtgdnJrNqcf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rg8C9jzgKnpdaa2vdBY8KM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E65A3eydyUZAh7GTceJcKv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VMcCt8EZxkPEkcfn5qMiyK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-94j3H7ppfLtQUiAYp7YHz9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BaV4DqtvWxNxyvA18NNpes"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3zsZBJAhqEkPVqFv4bHYnH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EWDrjq4ByxNwA6m4XQnJxp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-84GRYiWjq1gLz5P4deYctt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CGnB6hy82vmFvJgoJsLKPG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9aannnd6WP7QyKrjrvT8XD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7Uxyo38dw6Ggwb7BpfVxnC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UPQ6NMo62cBspiGq5K5ePn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QF4rjPpE6jcpvH1pqUAKY3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Neufk33FUnqg6sjoY1jnYG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5m2XDw5Lx31H8ZcgLRVAJY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Awwr55RnNecB9UvRnDy4em"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ACCU4rLBoR54CstbGeJu9f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F1ZKLKA9K7noeSSUg6rr51"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4qHCrUtXRhy1hBE3sbtEfn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CZ9DqjMw6wpNjTuiQ1Mdug"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WsX22RZHGabJZod8fa83Rs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HwkhUYX7R3rChWo4EkHGwC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CyCYF3GAZPr8YY5WiX4Lmk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9mK2fApZBqwFzi3StaVC1q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HJ6LSSBgndtdmQqD4v4sBy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RFHVoZR8aRqZyFhmL2Kdbv"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UPUJVxSEczorYyJXcqUiTy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ATnytnjrrYdU8njB1k2K3x"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1CrqQmSXKJRf1xGrR72MoS"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UhT4xzJiM7VSBCnvC9Dpuw"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1jzBPrKXRmgyd3g6Sghpyp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U8RNR8abGV5aH5jb98VCiF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K4yDQN2S7YNULCZrTBxmn3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hv44FcT11DBDAsyBsjJ2rh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VpdtU62t89QWftx3cnwWDw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01539_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GKCUyh15gxYHFoDyL5WqF4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HCXFSNZVzoByjVLm88GH5F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BjWR5VLrdnukfWPmGJoJh2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L3BTWtzE9jhvsXvYGL3EHv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32CV2MWYdrt6LormpGzGJi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EJwgPd7ZGgZxsA22DHH3je"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HjMLxYCooAub81n6NPzXdw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CsYCaeihvSpDP43jEoubpF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PFTjkjh4grQ892hYiDfSkj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PHKapuWyCJ8z479VwV4Mbp"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PrLjzBgbKVjxWXsc7onLaF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7294uBLkUv1SAiBZoFq3eN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X95TVZ6HFyn8hABvuiJQn2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M3yhvWBWQBkixbiBDFDcHW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NAPCHkuKSY3SBhXcLkzjR4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xe7xEGTxCrmWY2yDP8gCV9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QHZuSFvQ9DY7qsmy4JK7kq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1msru8RFvGi9x5FUPaMPNY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1VJeP6cpsbVKqQY9Hunz9c"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AjR6D6HFwkj2BcfRQ8mG12"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JDSCmGHq8SQUNi8yw4D9Ha"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRq3DgRzJYHLuaxMJYpdei"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TZ68bqvp5Y57kjkYav23Kq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKTyRCkz3fgbFsvy46FMrY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6u67PBDALoppXJDj94jd4t"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TvfpJkwuPABtDz5gPjBUto"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A9wz7VYupnntc4TrNsZ2QY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RVaboRcfLhr1XUHvLjHndi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q7zbmndbyZMRfNZhYF5DoQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QGTmTE2YgG3cChwRjmEyBW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L57qWH6aNT1aYfefAnvfaZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FyFpJJ4xeGWdPtNi8vzgAp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y01539_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FCPaFRiDhvCjXLWY5zEiPa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NgVfqt9GpWSCMjXfBPzuLx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CNUYUqWyeefcDXko5ZCtsf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TejheUzwDtdPdQLYdYPjQK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-28Ab5zRsdRd6Ynd7Ah7JsT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XvFyXda1dCD78GFTsXwTTK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKVvW38tLkqdQVKbwnaTpV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MH4nHPRoR99hB9YC6kK1ZX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M7pRH2dprgNPFhYXV5CSEn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RD452smHp4LhuTiNB9fj9j"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6tFqU6LSiHRUC85iV5DTZw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-45UwpeS2n1J5k4iNCV29SR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3gM6pcqKAUxFJYLwKxvus8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FRA9yw7sRE8yoNFgWZq1ZM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9ZweSrpv25VzqHX4GTwctd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U3GwpzoLP8u7263jcVDD8j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WDNxB6xiugJx8dsWsmJ2aG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DBoc3v7cWVKnwzZNEBcFbs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MX241C7voDL4iotogcLFY7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9LxZoDW4PrGLswXV3AySNL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3fhX5YjeiBTojuGc7dRcMu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BqTJnxgU1rkMoHmUN2m91t"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q852SUzoSrAJfadti9tkHM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NXV2CWEe6cHWdb3EwT5oh2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RKYtnWvWejkpHwKBNQBkWy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QiMzfArUajLLCXESNHhxP8"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K4hFcEfgdYPjnP6U4YukuT"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LJ8rNcyGrFZnWVL8to6Q6F"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VUDkxJuj4Xq4UAduFDwDmi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6MMCdTGP3S5tN8JJAYxfqj"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6TyCGkGaUiWEzPd628rsdo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3f1MNG3zC4fZCs9L72Q16w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LMshK7deTSn5dbMpYHaxVF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HSEp4kEPRKyFdYrahTqHsE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NidQxQDVLY18UZBtXuAtKq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EJBiDjb16is6xrU1SJ18qx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EMsMdpCYS7CNGNhtSemhzC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1YukQVWxRbsoGyaMNqekW5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7cNkutSMvhM4yHucpXUkKW"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xia9wTH9UDSR61A4M3rcoS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XhQfe42wELVupUJ5ertxfc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CVgzemc5s8nFEpfuvKkjKS"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PEL5tn8e7WpzH9QLfiG6km"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RCQEfRPhYYoJjWX2pnMACn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BpDQYBxFqA5m5S6vX3PBDb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FDNYY4Dn7C8f4fbMfGwa9T"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BZ8wfYYYnFGSJWDFdnxqy3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQh3uQceJtbn74aGCgAHZy"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X8ktQh8HcXnz1BNwPAyzDK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GUEjLqRJFQtabuND8LJGCQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2cii9AVeevfG9GR5FGJxTN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FdHSbssHCMF2LdrRKpU4m4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DHnrE6QaL6C9aD1uDn3HdF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7WxJiXTrf2bmMUXyQbovgL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Bxw7XqbvSNRCYCjARPGWvZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4XMHGXLd5HRQvcYPWMMWin"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CJpF7aFEmw4tsyT7rQKGpE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G2H13YmkmL9zftUankrqNK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9cUjjDDcJnJSabBHdgP4wS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LXVDJ2xK3WZ2TMPGHVSwNj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BK5scyMyH8tcxqqEvfgZa5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rn6RGqnsLd9s6Ujx8oxf4D"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3swq6NqCEH83eTjNr5NUFK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QPKDXeRfKkkm8cVtbJsGsr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XSXWGLzqSqcH96RNveDJQR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RQ2NzCqoNwcx8s9caGbed5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1SFZgFCwCcZLNHfVkdvtpq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MDXqpZ33EhbVRDqqKdDYep"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VG8jMC1hT5SrAxGLMXTSPc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N6uEA1HnKJc4PMeNDcCct5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Usef7Cf97sFJ61Gbo4TUbc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7bkuQjqAMv4pQix2kbwkGc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FEpEyLv5nnQUsGhTJ4EBA2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jz5Kf3RuF1j3PZd6Pnawfe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7PT1nWWCSu53PeWJthofNw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QpseBS1M9DUHuh34rMWhv2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7v3sxkRATCowUVKNhXtcPx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CmBVbok2myynpC1DxZ9jrw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BJKwYPf286J4bWxtnLq5sc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AHd4eRy4paiZFF1L1YQuuZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KXpoPAuHpwjbXhrxeHmtMt"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5cXth74S9T6QrKEbJyiKh3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HpQ75NLWcAy1PLvFALs266"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q8vBbMhkSe4ySvy4beKiTm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Azkzwckvtbf9jeHzCUYMJk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y00000_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BZDzRdBhXXFMdZSTdknAa3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-954FpkTrxVLcRrUMV2XUvG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ELRkQRMinSBTD1GxrhqStj"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WiDovQHdJXWhShvJpFpv2f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PShAiGp6UN7DUxHwCfP5dn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WyHzSmXvyJTAHo9SfTJTYm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F4tjf5C9kEHManPV6MPjnW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFL1TLmWqSJp2dMozVvCKJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FjaAYPQos5WoFbgZeZmooc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PpK23STWNxE96oqZw5y82z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JdLAwkFkYe1injfkuBaJPR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MXbzfktPXDeUTSbAxnfXh3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XG4pPqXjQL32MvVCF7qUyE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-16mtJLWuRmtaz4A6z83g32"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1HnMMgRvwLoZATeZcFvsdv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YYVMQMwysidNZjBXbdxsvU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HFZC57qQ58HVQkJM8zXb4f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-THVxyYVQEeYqoVAPnxR3T6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8VRWp8jK7idqKjBLF2s27Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8uZedP47vZJjj6jSXKaEos"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BR8foFTete9UGF7EVcfypW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EmWJnPo9L25zBVrCCAjE6r"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8YxYsyTiLdD5pNHbHPiqEA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4etp3Ke1XFHWsn1cSwVRqr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XAo1BZ7F7YjF3CXYgj8RcX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ef3SSQwdX1XEDyAdbMwPBs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04617_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2YbZ7CMaRZMSVt5x985ksL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A28ZsL4VdP94VsKSRkFPFw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2L3jpu9WPwUAGoerou2b4v"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RueXHrgsETWcEpYKcNZS7C"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y06669_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YMxtKK3AAu1RSbHh5YV9Es"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EdcaHCQTfYnKuBW7eAj59J"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4KMHUiQ6UZUxgaPpYgzbRS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WJuooXhapdUM7sTsf5Bapb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QpuubfHepiYjsJj1e9f7Kz"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7efpk2jzRAB9UwWYkezfCP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SVPVSrHQXo6LxxYqWkjRtn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y06669_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SUH5ZWqD3x2rTzkH2NEGcJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EoYsLSDuUEC4YnxLsE9FBi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RJ5buMmAd6NzbuCW2CM4Ln"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WKAQTgyWAdzQajfEJP6qwg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LC1aPfCpRhkXJ7s35oQ8fG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SCLLBcytoM5uHPBNAip53F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H3V4oa69dc3RfTSDJzwCMM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7sEUw4TUv9TKNdRwHvE3aY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NfPD1eh27zC2os5oFawXMC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EYMEuLoqKLmqEboKCt2jy5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8PXqqqXN6gaYhaCFYprE3y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FYTjAuvPZrys7TMe4qLZYG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PM7NJHhS7GmLtaFxeX3CLb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MZD9a6BXV4gAThxi6HP6nB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HYexWMp6EVKYU4KSGfd18K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RZcjRgNRxerLei8FRogvgU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8iuaUNVjhu9xp5C6ob9oJL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UhcbxNz8XFD7Vd38BRie76"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XS91uJDftbY9HD8DQV676B"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8wgXDrxRtsossUpCq8p3RV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NHgovdfusRnfFuH87MAqcP"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MuBXfRNBxAvcxRREJdfXk3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QVCiSaBPHqd46yVuJE3eEy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WjeGNiLqx7ZyvXsLnuScuG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U4JHoomwBVQtvDYXth4VRS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QnrcusiVz1g2wGQbEjqbBe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DJCPoKtAPyBZRKirWqwwCj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JCU4DpFM7kyVG2QN1FQ6jD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-96Qs6HdFfdK5gMWVVybLkS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WnfncATTBa7adygKzBcTMV"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1kyxF6CjT8TLSztJ6br33z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ERibD6xFtTCsBV75pnJpe5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8tor3gbhM1DDLssQQfnUU7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BZ49q9dtidL1Lj8tf8xUv8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VAu6Wops4pFs7HmoqQpVhX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PMpYBAxpsJvuCHG8xAK6CD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W26wiQcfMSt1azQGti59Fa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MQsMGDANAyBYnptvDFwKR4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1qxZjNnA5qgXHoExQpf8sF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M63h4Y8oKhxcqafzPPUUKD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EDVFrpqZSkDdoy1tEPWSi3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X41CkopSvLuavbJ4tvf27c"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1KqxMRHsuKuyCWJroTttGm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2s1fXUsqsQ4GWxksdV2BFk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XUwZojnVcSjWyAvcpcygg1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TJ3JMzm3kk9uH2uAXbcfRQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y06669_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LQ2VZwHQyAvh57qCmbWK9T"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M3MQUctUQAFa1GiaG9kNoa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LKxBhy88rFfABSBfZRSBMJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KTsFji8XEqVAZ8rixDpCBB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8iEsyz43VHGrPGhKGMiuBD"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YPgkoP2anqVE9d6TBy2K8J"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-89HSpWeE6n1G41hfSr2ki6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N9NKqthigGGsSCVBeXxa6E"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8XAoKjnhefT92zdkTtvUAZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GSnCjQyBH77Bxf8DBUKkk5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UAoqR9vzrFFgkMBSL4P7W3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799009-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N3DiAMbXV1QSErPdFz5ete"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AgFoMNcca1gDCogw8KpWD2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S3sbvvoVVi1CxLSRWC9vpN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BdeNRxuDok964ZF1vUASj3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VAsJPheBFXyGRuvUv9QdTC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-F2efGF2Wninmomr3pFgagF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6o5DQ8wgTbkZ43gM3GLFPM"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JVYwwukLDUfN377k2YCHMw"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1n53G1H4WZeHrXvmfPo5oX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HtZnxmQaCCx6yyRGjuGaFb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ELg9EvCwWF48GScKoe4i3o"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CeAs2j2MaoW78Gnoz9PP12"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1e18PLxP1pYZNQpJmUNYVs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Pj724rX8VWMuhFwFrGnL6n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Qgvj5HXjp3gLxnQjxXZC7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XTJsmqrbmJ4WWps8UvzAES"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SnT8vGbDS6dNajj1HXXU4c"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WqhL5iUoTSyrMCsk3s7xw6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CCtkEZmEWuoy41Gk6pGhU6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SQR4RFx8sgRBVdpSwPBjhh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P9MrXTJXHLcQv7PJWjYSZ3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J2Jg75oUoe9agxgEBy6z8C"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Fd9fWMqh3PTK955T4RN28"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HsUXjubkSR36u4QbU8exed"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRvDemjr2uHFsAj3qENXL5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2SB73iJrdAgNAMn8x3uJPs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QZkYVmBeP1H7PoPRtdCbub"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5BS4Y57j82gKkpGfLrGW14"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SxdJGWrkQnmqZfUVEstsFe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FvoNNcUrStpftgYvxjikFd"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PvQ5XB2zmidZ7xEVmmcA2H"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RMGvQMrzX3tLpP7Rm1FR1y"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7jhTeGmUq4BXG3E9RHuEnx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NxZZAz1MGR1RGZLhUpga1j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MU3F3v9NEyydZoEo8RGU67"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FgQZ61So8TqLm9HFfFzynG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05643_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Mt2rtvnKeqk4qkiExf3oX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KpZXtDSowduTjM5XzqhTVK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5niqc2GxnHWA5kBZ2LoMQZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VN8SdQ1pMbm9TM8tChsHsM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001767953-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8vL6dWtvnKPdrRtzLhUMsg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UUiGrNDFAJojYnX1x24YaM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5R4Qpjus7JbZjJr56M8rsi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DLAR9uyD85E8mmnzghxg1W"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XsHvuVbtxdrbbdZB2P8YJq"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eqh2moXgcigdDJ5JnEKwxC"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1wKij41UU6uvqgS2tL4isA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8wTpy2CgrTof7oCniiYhqH"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CpzBPT3cXg8dNu2UsUZVKD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vwe2MQ8sqs8mw4SVP7tFL9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6mJtkMCFTK5nFbbW4buM8F"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GvURMZbkpYuCX8thYwee6S"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Uqpifer2MJ8R3Jx7gEELW6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L2wUp8svgKyUkRhExaWYSt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2meve8mV1X86DjTWdsUFPh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JKyE27LwoaF6Ts7muRM7BK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KBt6Mwc5DTHB5UZkZJMoso"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XfwbB3KM4myiS87oAgb3Kb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UkYJVYFQz4KKZh6urYASTU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PCPoyRncRCskH9p814CPm1"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MGSkJPF4gwXVYf9c7TrHoL"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TvhtudDTpZxyCKG5iuS5X7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1XyGAJon2nx5sdtyjvbUym"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HBimHXqhJPNpdsM8fohcP6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001886305-MSS2_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VAdrNkq94pyLzvYHMVfFfj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GfeQMygSGjqjJ1cSKkjfdp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HPy85omwCU6pU9naE8boup"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y06669_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SqZYEeXcnymtzhMdC3oBbb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NztCnqJsnRmTH5EQFcCukS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05130_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6QAe1RTpLzuZ9JgtuL6oL3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TRug3Q8D66wY2zTiu8dHWi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Psp11mxQuG3DMcM3hQNDk9"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y03078_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E5RCUSDrw13LLrJNWdU3Qh"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XNyqh3rngkz4MzmdXrKR72"}]}]}} {"custom_id": "GF2_PMS1__L1A0000962382-MSS1_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xz9KU9Xu5prneUPohBHjA3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LcD5ZALGr8QzzRxsM9Q8u6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6b1A8LWPcMXrijN4hyWcun"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sw33BVbJZiXLwEwsh2qiuq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RBzD68ab33jrKJsZN3QV6f"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AdPVMKzxJnPj99nY7QB8bA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Hf7JVhzLhJyoXj5asxRX2f"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D4NvijAjVRf43UvCwW3yTW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Eh4UcMT1QkMTmymqUdJTxk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566654-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y7TqpTTVv39tKJyzPboPBQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XU2tEBtsWsUpCbd3xVmaMv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-57RANkJ8NJhaSN4fpobedr"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LiWMXbd1mDpSPYvRmor4ie"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Botb9Qxmt28QajZ7sQbwps"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7n675vnsttGXgsrTWbxMPM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-48MBka88kxC2S4sc1DEY1S"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7CHSbbd6HnCtZAj7FmbaKc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WxrrEokqV7J8MCoTWG58EG"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QFFtuufzRrQ4BQWGUJo8Cf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D8t6UXAg2VuhihDrAbhVwC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MzBjWqo78cpJpoc8Cfaujp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6KYKRWwdJyXhwyyhcrk1Jr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nf58MKYeexSBbCvfbjUPvu"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1B1iTQTKP6rs7pPSSo2uyx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9s23DXNGnZFpo8bHVQ7MDv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RHx6EM72k9Lg3wow3jLdT6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8SFJPbmHVHSkfNAgC6hWvC"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G8tAsn6WttkzwUvQEU6fVs"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KDnLCnk6sazoprNJeEWsSX"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9A1Ntu66R8M6taGSd7Up7D"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y02052_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WoKrFgQ541rHy8sDc6JVGR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AmMudRFNjbvwPRaexnTYng"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DFeHD5bWqndd2qN1rxZc9T"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VZsfaNcCKiQrBogfvXXUBJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JY5ExHyKcqzv9ATdhzikYe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y03591_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LJT5YQxWoMSUkgt47CARLS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7aRn3MJtRSBavLnN4FbgEF"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DHqKZtVVnXg5ErQKgrp2tv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y03078_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1iUo5NpbZ4Rzfkj9yUeTPC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TuxqysbQ4dbojGnVeHjPd9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JMS7aoNAzTfGZr9PfqUSc3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CbrEAeFZFBmTF4LWRwE5Ey"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PeBmVHtUr9xTuKcm3y9LcU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PjcW1neot7NQ2kDdkENAck"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SdYPeGvzF6Ph3VNs2VsZas"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y02565_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Q2QtTRjzHdBoiBytm8MizD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001910522-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ERSJrhgFREzuFCRtoKoRko"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nx34PjQkpBQ8jaswMFjowZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MWeD3DNKgaT4F5RFQHv8Zd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3D9fgyNkXe8HMkYEaUE8gm"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JLknG2jRfqoauecvGy211w"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8GYjzGFHhNGFHLBBAH3b8e"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Enu1yTEm512XLZcCMNkwwH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WhLVUxTzxnuHie79kfQ3QD"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y06669_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ALNRMN7KdyDXH9i4Ddeq2W"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-52uLs8qDQDAU9GZybn3yir"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5zAswC2j91mfUDNBSsVvFN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GcDF3bmX5d8UsxCBbxF9oC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DKegXTVzZEcAsX5eiTvjZb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRRLRQXi5VxiAT136zsUue"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ut38jC1mD5jwByary8ZPoW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FN8nCjsDyCc2FHECwUofky"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723991-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B9FK9pnsY7dGLsw1UfW9Sw"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GjEepcequiBY1ACfYnqW8h"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B3dXSDspWHLqZAL3Y6yWBn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NudzifUSZqB3yXcrLn1uJU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RLAgpiq8eG59oHTFhTgmJ6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W1mtTqXAxhdNWTRHeQNTAz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CfheG1h21t4vDAF7SJoLuo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SuLX2iwj2XVrMzrg6zARBh"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GffKnAZNB2DgCt25wKhni8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SusqEvtxxQZViaZBNKgMKu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSy8EZA98Ee1qErmwAP11k"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6MJmFDCyRJ8r2JaBAedmgh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FwXnhnCdTRRE8mLPi4Kmji"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KnkwhgHYHoWvJ7BStGYGoX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001793003-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3Svm9hwkC4wpVsv6zv2ghq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W7wLhBW5hxMtczLNWqnu3P"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DsV639nSKLKM58cvseCBDi"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y05643_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XbZaSxupKEfefL6CWZtSVE"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3ijK7QkQ3uEkj2eEisR1D1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YQwfr5MdufWK2EejyxMFVg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NvPgeday4vSeVwM8oS7tmJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06669_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B9jCxp7Y2XWJb9T2KEFeLs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001389317-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8YrNqqb1Zwb6TEodgpVjFN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WMWG232TCPQkSK7bjFRPj5"}]}]}} {"custom_id": "GF2_PMS2__L1A0001471436-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Y9YLYWEG6rYWt5aPK7UZY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6NFJQowkuCX5xESzxghzBg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B9wywJWPK7Et57v3nEwTPZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XfGMFD6j9nmKw5QBt8eWTH"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E1DmK1UaQ7DuqARHCm8Nsv"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MoqgbeqrovWDBk6WEjhFfx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DoTpy6rBR2SMdrEmTLB5ec"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P6K2J8Gs7W3ATWpwNfz35s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y03078_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7sBX97px5Qyg6QtJqPa7id"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RmS7PmKMccskQ7kYHKWdS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6vxLmnchnzZTBuNzy93wy9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y04104_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GzFFUriZ3WS5fsgXsCzQs9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QzYWcQ458N4kXwRPFzSHpF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ESdvcNE11f2rt7nVxwAk1G"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1WzTh7s9yzC2AWSQaqytGx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WHUL7mF8xr65GH9CfMJRZe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8crUcmf37Rq2cndfxtmDBD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NMkxATgfprxh5d7S3C1iNx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7qXUjk2k4qLuybBYUgp9xB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378501-MSS2_y00000_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HcSbfjT5vff9QLX6YFdfn6"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y05130_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4FuQ4iAdKsqfTKiweKyjQD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7rLkfy1M2c7wSzaTBiUmQQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4o35xwzrVkwyYXmYBDHVXx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y04617_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YN5BqGPWaNgzha4nEUWHH3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73j9LbURC1FVb2uqQSu3xK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WgeDuDy34Q3em74NMBi9BM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BQqUVpujL9BYwyLN7xXRoM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NV3pXTH8UTF9U7D71yh1iL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708232-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CwPd1WQnGGPz2gaV6uC9Ad"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MVPwo2pFp44T86Y6XR4GzR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6givf6miaK6CA2xA3ZBJjE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4J31vR3cSsyUFqJdHN5DLe"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L18YRdnNr442HNuCvCUr1M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001746797-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PSPWEHGEwdpGPnR6FN4NLd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JGFiYknv6r9D1BDgGZn5ed"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-34dtsZY2LmJSeyUEd1fNbQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5jfMjgtzXvNCsH6VRqNoQQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DVNoNFdTXS2wzbiiNaVB4Z"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y02565_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y2hqtv7WNJ2cvnrobHAgr9"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y05643_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3miJyqjpAMr9iVRHg54Hy8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MhA2Fr6K8r18tL5Emh9wJe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A2fvZv3Y5rshYtR4Dxrw3r"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y05130_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lyv4XiqVTmP6bdfpNNdWaD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7HUqG8kG1BC716r22knFyw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y9FQzDJ4CjmE1tSdqx49Ki"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fno1z94T1bVrswyhQ11tPF"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y03078_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AcoW25PgBmDbZGSwCwHq3c"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-43k99bcFdWd2i3jk4US97M"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-89NLSSbKsuDqC1bAve3TcF"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y03591_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VbeG1JBeKuHqWJaLWheGky"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y03078_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B51D5wriHrMixa51okGUre"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J7qig1iqsqrRUjqUBttmv3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QcAtUuSAzKT6wXDjsCqMzt"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4UdMW3o3exbxrp4vgJLxrX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y00513_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4ZRpCiTQyNoKczgNMiXcxy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y03591_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2WPDqNF7QvZeXkwBuNuRu4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3jbSeZUZh1DNurD8rByTtQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TaE956RR65t3Zg72HnX1gQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y04104_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NUGrqmw3DmDeRPTSfC4Ajg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BhTQM9dqTX4sTW56BAQnpF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y00513_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-62rtmvqquk395b5KgrkMrt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3v69x2jtrpUKnZcbaa6LYZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G9hA29nEC3a1fRuSEkpzAJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WchJEc65B6Po6TBCcZ12GC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7hQWrNvJi1NeQ32XrTbviN"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4PXfaTecDfPjtZt5hnbQXR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EqJbDt99X1FsxDuSdWxezQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9yzZPGxGSsZ3cAvkFzwEqx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-866HpSc8z1qvKipULgcDRw"}]}]}} {"custom_id": "GF2_PMS1__L1A0001118839-MSS1_y04617_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1q94aRB24UpxpnBB8WxeFh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G9HmrV79ws6hnviorpyye6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7NXCtpULw9mrqZbyTQhVjG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DdkDJyK9DtGrRuLDJgF8Qj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SruYy7g56aDLcEh1CUcRHr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00513_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SQLaaDxXCShD7UTkuBcXzx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QFvejxmqcC4GA4vDJTpFbK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y06669_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MXde6kDgsfPxQYTd57qEX8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tjzo28EbcZzyzJ83JgvPtP"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Xa6RbwAHxjTemx4fX3NZTN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6YXxnrns4AfecyWUjCAipA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PQKFuKq3nbApu797U4y5jh"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EqJo67t69vSN5wvbBHcieX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001290139-MSS1_y02565_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HbSetrzKLgwaCHaiCAu5rb"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416129-MSS2_y06669_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BzTz5f1LKhTMQG8d3nHGiz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CnARsAG5NquXcRzKuPbiWy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y03591_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X1zpUB4qbkojHAk76A7cZc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5ReGEfz2X3tz6KNuMhTxid"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Go2K7i5eceVqfzkLqWcfhs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y06669_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7R6vdMEXvZHAMc9nAFsxmb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491484-MSS1_y01539_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VCoU1pukCFc7aA1mkGZLyK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QFAFLzp1SM5VwEeJdVhQdJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FrtdgCExK4k2X8tMUXqd2n"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LhcVkQLPS5XjPUcNh6beVp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-48U3FMPMpPoNLbUouHLziw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RDNPeMhcSTyx7HgTfFKx5p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y00513_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NbEjYCHQmcQi1RakXAsy4s"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06156_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-G9z3EzKAErBtdBBDY67F7p"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7f1GQCQD1wfVHcZP5Aahkn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QhhJE3wWaTovPCbrBwGBGE"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y00000_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CyZKRekQ9WzLBSucKStRrS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787089-MSS2_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YTwiDemsbFaEqxgk8PTrK4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QSQmSWgJ6fwE9a7TKofpX4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y04617_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LRsBzxcQPdfBGhrXs6q8wy"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-42wxvoigFJu6MBB8ybgop9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001668457-MSS2_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1hxv59rCc7jmgNvFCJsmfQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KaP9Ev6ESRmXE7MQaL8npk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001448894-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XQ3W2iG6ytuGcNENLqLrLu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y03078_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1DrcdNiZZDBB3zxtaad8XW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001378491-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7XmqwCYeafThPv5UB9Qxmp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y02565_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HEsZLbj96DaEMJTW5kXcSn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5Ec92r3tTe9xHenQg31uzY"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765570-MSS1_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8cwhtjH2jzfZx465Wt39WY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QfFL93LfGBG6yzcoqem5wo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AzeoMei7hzcmiR3Rxqq7HB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116444-MSS2_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KSbSX79omN6McGf6Cc8S6z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JjmZGAn481S1yuCNwYJ338"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LKx7JGogHTgXPHPe1dG9P3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2h4hzNgta4zBh1swHHEtBC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000948183-MSS2_y05130_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VUZZkiBYRVWmfKTK2EbptD"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3P3pq2sSYshQcjWxC3dpTJ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y05643_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B1L8TnQQRUVZHqZAHspd5k"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6u7ezd2hgeibJMKHWYUSfn"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3K164ukDzpjmtyJ47gdyGL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Jes4N4FXkmDvaJST8wumtq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8MXTAEH7DzqHvJqUHSnhjo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958144-MSS2_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-L5MG1PpR3oD5KuZVAvgJpA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XqfDFaw1XjZu64vivUmaLj"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2abL6PZkaH75czXfRXHnpu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y02565_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WLo24iK745eZS1JJf8QE54"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y03078_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XLcYScW5D6DzGtH1Mj4sMT"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CL79hsGJgdkaeWysSwU7PV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-K9P9MPKrre1Ytyranoh5Xv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-43PTrWYpLfT4SkR2uxzEGe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TwvcRnKSYSv8fh893NSJvf"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fcwo9LvvWQkRBD4cNtZfud"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-74wrXN9z9at4JtyoW3gxMK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y01026_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X6r6wDBDg21gb5WVxszd64"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2uGTwfpgFaLPuvappZoTUp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y05130_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QHJsfvbnFfa4xRwnREC7bW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WQnXeAS61LvAeqZPvkG7mp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QsRY39vEEae3BRRHzgnJ1e"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SiAedsjqfVoQUfBXXExZ6n"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RmMC1iUFTKauzATFQyJZ7c"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APS2tZFeSiGChUAZsA2WFV"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y02565_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HMg4p72bLYwisapgXuvWjc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787564-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2id94hW2rkNyYpCQce8Wkp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YMDiwZU4TZ1YWsV85h6WRi"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y06156_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1Wo36ASDnn4EbvCGENXqsF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FFNFxNg8dfdwdGG7EdcGzS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01539_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XAao5AnnrbrHUD756vQAqB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y01026_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3o1BHPKwp5mVi2rNuBhYVf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1zYrMPoiiEgYgDo3ArUPJG"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2tLE8hzhuJd6vf8MJAoM7d"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537636-MSS2_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2RrM3UM63pAKQZkCMuNnor"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DYvpY68WjXKYFdgbhhc8mE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06156_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WpNxLxUZtyBFLQmtqCHbtx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MJbjxwgLHnzyZjLbdRKevd"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y00000_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D6qHYYYr84sgKvd9pKxCn5"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y01026_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WZyuJddF23twzNgAfCBBWS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KtVnw2pq31bRn6Jy7XBcXc"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8qwzxt6fbwu6GPgbKevwyw"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-18vqauDPkVf2WfxMYCfnM2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UaY4giMmr6FhT6An2qiAJ2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6JSmfosYN7nLrNijTGqHrM"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y06156_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5RNCJfmWNv4fMrPSpNauTs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7VfkAG5FKGhHrjqNYfyTYC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UPxJaYgJXsmUtrd7YMuKVq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680851-MSS1_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y1L7Bs7994hDbKmGkVd7Gu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00000_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-U9WmdhfNXC2vSqXb3Ln9Ls"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fb2TBSKmZGXXW5SGAKH5qQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JUiAAAxAgrxzBDKghbWe4N"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KV5mXpDoXSPviF3E1RYJ7p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y00513_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9vp4sN1GaHGb2MbJrY25wu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y02052_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YJgf55LdGVduTXpXhVK9F6"}]}]}} {"custom_id": "GF2_PMS2__L1A0000718813-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LydLy9vbUMuwEq3NPp7mxB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001335463-MSS1_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GXGpcsfxinhWYzjhTJdS5A"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-AgyesNSs91xeb5nGsxDqdc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C59Vxj3251DbRzNYgsM1DJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001838560-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4U2sKAEDXmzjooDT4H2WDY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NUoofNDKjYyuptiJPzUEz5"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PxBemW1i5jEaDZVpuLDZ2w"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y01539_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-T61FjhqvBEgEjMJoMoaVMC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MLyK1XTovcy2qyQV3dJZGc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y04617_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2GdwLdf4AmPbEGbmfmDL6Y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001471230-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GKEKvhoNa2dtk39umKJhsW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-D9vKes4QibdRcXi591nsXG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y04617_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1NZB6AL2ZuAQnXbxMq73De"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05643_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FqV2gC8Hm91bsQr1jDQEj2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-73NHGqTrHwHP2XDeq4YFov"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H1cynGS8WpaT4H3ogpmAjX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QymynKLhj2dxQLd76vuWcW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MnZiGpoJoqHYDcMVEYnfUY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2KB7wxcwAWN5C8ET3pfXkJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PZFDF8QiZUU72rbEDFdeVk"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658636-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KqCfz4otaVs2VKYaRLJYY1"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647770-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J92tsRU1aGydytDESCEJHC"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LdRcCYFpQ22XqrD1LH4Jxh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y00000_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KWfFG6xoNcu3dmUGB8A9QY"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y03078_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PqwUDS49tWdJuQujyb962T"}]}]}} {"custom_id": "GF2_PMS1__L1A0001517689-MSS1_y06156_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-78ZFXvdDy5acneic11Dnvk"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y01026_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y62JWL2UqoX36NRavQvW7N"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XpXs24qnH5AzcXpswrmSFb"}]}]}} {"custom_id": "GF2_PMS1__L1A0001348919-MSS1_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QqadZgzxaJWvu8SRW8DovD"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y02052_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CTmpfCY7Um6VWKgJydp45i"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015597-MSS2_y01026_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SrCjtSA42pKK6ySPtPVobe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001553848-MSS1_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Kbo9C7YevemnWKLyDYrgfr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4bw6M2zCwFfWfBHrmfth8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4CnMjxkkzvnh6STTt2bkmy"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y01026_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W7rr5cJHKBfPqELv7XQSCU"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y06669_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C7Gk8RA6VzMTbN7EX6FyJQ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-663o2LuVAtNS4JbdgMDnjS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05643_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5FQdhzfutZZNUrSM4MSbc7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001821754-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8pnwMaWXUZKgeAAVLtMt6"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SKE4JXHcWswb7EReaFjGSo"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y05130_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8k8YvZ8j7DgB3C9KBQ1wbg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KD7Lp9dZYpXnJMHc9UHD1h"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y04617_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YaXbtBEM6Zwf7SGwioyB8J"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798945-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-P6F1ZNJeGoTuMLkox3SJYs"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015649-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CC4rjjqFjAPaqMm6tCqk4K"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y06669_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lvr7QtGV9hgyyxVvaTCoGK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4n6exhoWkTpnL1sQgJ1GmX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Fopru6nueqWaM39PZEoL33"}]}]}} {"custom_id": "GF2_PMS2__L1A0001433318-MSS2_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UQqYaMP3BPwSJxFTgzZWmt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PceMJYSR9w5p62TQBUW654"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y03078_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4aCGW9dUAfgotPNSaD3VQ7"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BkqAqh9TdQcArTk8oc7oae"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8iwZYuS9vFWwY8R2Pn9xod"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Y5wfAz7GXvgtthYwZfP1x2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y01539_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TfuPetLDdGG8Ti2PKgSdnt"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-W2vJG8uxMtiL9R9wUdLwrc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y05130_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PnTJ6j3ug6BpR8DRQM6Qf7"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366284-MSS1_y02052_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S5sMQd9zm94MqQAiQ4xBp8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ug5MAMtQi7ryTcNQSdmyCq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LZPQ6HkUfgdgFdsQdTU9P8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y01026_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YAEb3n2AaaUCJmoAvRwxLd"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y00000_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1cfYLrYLKR79xwmcQzC1D4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001393716-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6NwzdhR5Eaab5dsyxnkPro"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PVDE7okV7ZDh5Gt7XZNdQJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y05130_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-114xTMzzserNSkNxf5REyg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y04104_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5rBiHiDeDBhJ4YCsXUKorW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y03078_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PzZ1nkHQ8jU7xrdWNBhTyH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y04617_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Gq1UdH3gSB84EfPERGbkfz"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Dxgc7C1S2tZf1hWeSDb49u"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8Yrr4iNRCowZ3oqTAD3eyR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KLTeVqhrERpafSKyhXmvZm"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4mY95DPU25c6x3ag4at41M"}]}]}} {"custom_id": "GF2_PMS1__L1A0001660727-MSS1_y06669_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BRJHGFfakMsTS9TsbAnBAp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DYLYoKsau6rawGGq2ooM8p"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UGqtP8LKnwFmbGPTHhwSLG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y02565_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A65hHWxm7Z6dPdhq2Seuf8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y04104_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8M5iZwhiyJg2eYooEp3yAe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001562063-MSS1_y05643_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EJU4sBqPR58HQzSLnLoqcc"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E2N5acZF19LQWFCG9bAGcL"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y06669_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TuR4HGJXgGLjPWHr8CeCUJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000658637-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5YJLmZ6zbAd8SD156fsRcX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UVTDMqVaW2m2W6omvcx7jG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QNVtFYNKi9mgpcBYEcEWJJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y05643_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6RMJski25VV3SFs5uwK8EG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396040-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-766PR5zJUvhZm5VY9xep1R"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y00513_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rsa6kP1ewpaDSJrkRkfgsu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680857-MSS1_y04104_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GFZqxXCHPQNt6pYe64n7LE"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y00513_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UfMor7GCV6dtLsfXD4hcmc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Exabj5MVDhY1TYXsTiYXKR"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607677-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-44vWBxiDuxZWLSgYGgepu8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y06156_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-N8YYDdEVhWG3QcgHF2f8mx"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3WXgdVgouzdyxyC9s4b2h3"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-S92VQ8coZEmbe3nzZxwkQg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001028977-MSS2_y05643_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KrwhJ5BJ1r8NrbQhbBKcN8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9k7q2nUnmaTkRkRewms8rH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001037899-MSS1_y03078_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-J5SqFt7QN1UQVWnGyJcAmo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001094941-MSS1_y00513_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7jRB4sF3BaogMD1JgKtKFu"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SztCSsgaHSgaRBJHMKQCL4"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y06156_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SicpCyGHjGVRMM9NLcS2eF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787080-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-98wm38zVLHz1Dw2c6t9Egf"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05130_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-75iQyW1CNCXf82vxVa6sf6"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757483-MSS2_y04104_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-34FTte4B6szyN8BLABoZeN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757484-MSS2_y02565_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4FbDqjrtGnzAV3QQWqgE3d"}]}]}} {"custom_id": "GF2_PMS1__L1A0001734328-MSS1_y01026_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JFmTPm3c8KptPzWq7pocjS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001537716-MSS1_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3iKVhkUFUMzqdwiakrNEDs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708230-MSS2_y04104_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EYjBcA7k8LQLXzo63MLnzU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001642566-MSS1_y00000_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BwNNpQAxL2vN52MhxTFPqx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y02565_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5YdC2wHpDmsDxbSHVEoxdZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Wa6zYeJXrALJJ3itXg1EdS"}]}]}} {"custom_id": "GF2_PMS1__L1A0000575925-MSS1_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X9srb6uQqk4ZfFAMhvG6We"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-V7NfXQQmDoe2wx6ktaRwc8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491423-MSS1_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ebwfgnczn1zC2E3MyHYqga"}]}]}} {"custom_id": "GF2_PMS1__L1A0001217916-MSS1_y02565_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-M5STLHCrbJ35Dda7C83UMB"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Li6zPasR6EzdZmaUrymg27"}]}]}} {"custom_id": "GF2_PMS1__L1A0001395956-MSS1_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-35Z9ix329C8tgdgkX9xRxe"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y03591_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EjXmb5DjRDSNYWp99Y7LuK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y05130_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KiGjad3eiBjRhkNNsP1ihg"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EZ3RJxmiqcMDfCPaRTBQDB"}]}]}} {"custom_id": "GF2_PMS2__L1A0001876775-MSS2_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HQucm3nmE7mfccNrrkkwLC"}]}]}} {"custom_id": "GF2_PMS2__L1A0000958146-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4HKsnKi4an9LGCdv384Uoi"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y06156_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PdQkn4s215dwabaFpNK8wx"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y00000_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KDb6fLuFysb5jgfVuBMqUY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RyQhQ7A4wtCc3P83curA9A"}]}]}} {"custom_id": "GF2_PMS1__L1A0001821711-MSS1_y01539_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SukGf3BMEyxu1wakwtNvWc"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-6zn7xpX5MCXmMcv9AG6DFz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LTKVdoL7TtdVgXyq3AzQpQ"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y02565_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TS4c9TTceLRM55HtVoS2Ar"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GmiNGf2fXj3nthQdr4dKjv"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HpZ2HRzcFicsG43WdgkNut"}]}]}} {"custom_id": "GF2_PMS2__L1A0001577567-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MYo6Mkem62Jvsk7TU5PNGq"}]}]}} {"custom_id": "GF2_PMS1__L1A0001757429-MSS1_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UfpNexVZDuF1dYfkjhQ2jJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y04104_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4JKfYT2gYdcfwxrwzXg768"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y02052_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ATTRdaUYiyDCMmPMrk7FK3"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Tr4DXGbEmU6iTZSts5nERc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3rNPwSfQ4dQdppLF5j7gR8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y02565_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JubGz9AdspH6uCAdHWpC9J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y05130_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XkoKvmmjEzN41Fh47LqLua"}]}]}} {"custom_id": "GF2_PMS2__L1A0001573999-MSS2_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-C4x11CaStVCZyB78AaTZc3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001104323-MSS1_y01026_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NsBzLXJh9mDpcgegmNvX6D"}]}]}} {"custom_id": "GF2_PMS1__L1A0001798942-MSS1_y04617_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-92gNpxgs3S4jXRmdxPU93d"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647768-MSS1_y04617_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1AqKTqZwuByR8N2zauQdoq"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FNqFsRdtybvWDBcRGRR6Vp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-JckYosFkQt4Z4g6KM5yoeX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799010-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4Np2z3dKYQBj5b5abwQ9PN"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BhL8HqMmcg48kxnqDwdD6z"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y02052_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UWEbCx1J2mNhWPnS5dWKrR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001824765-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KRbe6j2JCQ91BKEbCyeyKe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000736601-MSS2_y05130_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-KYQs5xodCDgAMp1EPH3xYe"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064454-MSS1_y06156_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LakHdQinswDMxDWoMM91VS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y00000_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DgiFjNPcQ6VcmK22iBDEN3"}]}]}} {"custom_id": "GF2_PMS1__L1A0001344822-MSS1_y01026_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GbH69B94i9XpveswbzkfEp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038936-MSS2_y05643_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PvG5whAEDXTKssuyt3hygh"}]}]}} {"custom_id": "GF2_PMS2__L1A0001768200-MSS2_y01539_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-A2cp5L3JJiEsK2kG7hdpnJ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Nab6aMK6RXzXY6SMDs128t"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ToBHhbCBeVLdzkUZYkwfKK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-79osQDKtCnVv46TFzW7ijg"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708259-MSS2_y04617_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ahi5V2MDnDUSnuqLDUDLFK"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y06156_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Ad7vdjJcGobsm8mSHRfATU"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5nDHN8LrUMcuhyc7LBoA7j"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119057-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3PhW5AVbngwTtR1DdtBRau"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y01026_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QpVHyLuE657bPgseTCLXai"}]}]}} {"custom_id": "GF2_PMS2__L1A0001140983-MSS2_y01026_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GiefiQy6W9pUYCLqtwU5fd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y02052_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1hTFFNwph17S5upa9TtGV8"}]}]}} {"custom_id": "GF2_PMS1__L1A0001293546-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XmY1VFqkV2emsnN7XioSkb"}]}]}} {"custom_id": "GF2_PMS2__L1A0000647892-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TfmZV2eCwaWjTiqyCcNDxs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y06156_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DVVnDMZCSzsRzxQHXp9pbe"}]}]}} {"custom_id": "GF2_PMS2__L1A0000788763-MSS2_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VFcLrisVdxqc6icfASDKTY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001517494-MSS2_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4WmwtkCk2LGPnMPWbLZBVC"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680853-MSS1_y02052_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Lfx1EsSixbt5yPKCLvscV8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001457157-MSS2_y05130_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9skebha3g73qEirvkfYHcd"}]}]}} {"custom_id": "GF2_PMS2__L1A0001708261-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EVp2rMMHg8jzzCdNkKz3ye"}]}]}} {"custom_id": "GF2_PMS2__L1A0001038935-MSS2_y04104_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GTYe839RWzhSoVdVvwnpgS"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YKYoXhi5B21As7geB9wUHY"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y01026_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJVf15GPXLCgfTcVoqGHm1"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00000_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1fs6HnVS4cHkb56nwFTU9g"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564691-MSS2_y01539_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3K6c1yV4exDResTJFUF1sc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ECpA42cb6hvDpctunUGEYx"}]}]}} {"custom_id": "GF2_PMS2__L1A0000851345-MSS2_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YRsgPdrxuH8hsXw3Y82zB9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001092725-MSS2_y04104_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5zgUWTbqEHXKz6engVqSVv"}]}]}} {"custom_id": "GF2_PMS2__L1A0000789259-MSS2_y03591_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HFm1y7bW4LUQmRdQ9BL3iz"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396037-MSS2_y01539_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2kPT5C7KwCT2HjuWtyjSvu"}]}]}} {"custom_id": "GF2_PMS2__L1A0001799015-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TmUUHz3Z63DNyqEsTMjRZX"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y06669_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VoKLxs4HdM38YSt5BUBc3a"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251799-MSS2_y06156_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-32JNuSxUUcoyskNz7ZKN3J"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y00000_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-PJBL12WopwbDTgFR9knzyL"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y00513_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FrkVakjk5WJNE8dwKTgDeW"}]}]}} {"custom_id": "GF2_PMS2__L1A0001633212-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GbpDiWHXftTg3hS8Pen25o"}]}]}} {"custom_id": "GF2_PMS1__L1A0001625023-MSS1_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GQW82QVN7fs9v2wavnKNJZ"}]}]}} {"custom_id": "GF2_PMS1__L1A0001169100-MSS1_y04617_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ToQdbfSTjXQF3cfGtiAxvU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015648-MSS1_y05643_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RmGkpN69rgk4uYwcRod8dU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XDxRtMzMho4fooYRkGDrfp"}]}]}} {"custom_id": "GF2_PMS2__L1A0000607681-MSS2_y04617_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Sp5L6gGW7AEtqY1eVePSV8"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y04104_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-VP2HQu5SbJLoaa8TJpLo5Q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246645-MSS2_y00513_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-NExDdabdSNn4AmKohuVN7k"}]}]}} {"custom_id": "GF2_PMS2__L1A0000635115-MSS2_y03078_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SjDW2kpQ8KziPeDoeq9ZRS"}]}]}} {"custom_id": "GF2_PMS1__L1A0000647767-MSS1_y00513_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CWQZf7hr6D1He9q5DQduDS"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y00513_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-E4RkgAbxmoorJcetNuyycc"}]}]}} {"custom_id": "GF2_PMS2__L1A0001416140-MSS2_y03591_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-1HXs8XfKeFrskMkwARSsEM"}]}]}} {"custom_id": "GF2_PMS2__L1A0001336975-MSS2_y00513_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-ShRTDjgA8sxzgWTUhwg1Ri"}]}]}} {"custom_id": "GF2_PMS2__L1A0001537637-MSS2_y00000_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9Wqk8pg9vtYJjHeVRh29VH"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y01539_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-TdD8NerSRVgWEuSgX1sdLk"}]}]}} {"custom_id": "GF2_PMS2__L1A0001723989-MSS2_y05643_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-62gAecHcVxRpGSeFKLhdvE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y03078_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-HXBV3ifRE2URwxq8eMLGbo"}]}]}} {"custom_id": "GF2_PMS2__L1A0001251800-MSS2_y05643_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4DQaAczmBrrqryzCzScLGK"}]}]}} {"custom_id": "GF2_PMS1__L1A0001791738-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-DqVK3aBZJo4F8aqmbSQamn"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015596-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-B8CJ2VBu5eVUMBvDXpmZNF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367842-MSS2_y02052_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Rfrpy9N2wRtp4bso5VHCVE"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LHa77Eb926QiA9xBawD69x"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y04617_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-CLJ3SP27mvCw6979matKui"}]}]}} {"custom_id": "GF2_PMS1__L1A0001792971-MSS1_y01539_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BAax5sPM8cbMTCNUztXhqa"}]}]}} {"custom_id": "GF2_PMS2__L1A0001246644-MSS2_y03591_x03591.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RyxACWCwzzFeYVbiPXLcQ9"}]}]}} {"custom_id": "GF2_PMS2__L1A0001591676-MSS2_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-WM4mFEW55GPAjE5ok3krYG"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396036-MSS2_y03591_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-5aUEgtezaqrSwLWsrgQpvo"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y01539_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FPJvJAmyj9D5vrj9FuJgQK"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y00513_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FmkGZFNFiYoaKpC542YbBp"}]}]}} {"custom_id": "GF2_PMS1__L1A0001112719-MSS1_y03591_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-GgsRtwoDzHvSWPhByeb4G4"}]}]}} {"custom_id": "GF2_PMS1__L1A0001491417-MSS1_y01026_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RLcAtDVBMnzfYPErnUsPfT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001432972-MSS1_y03591_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LStEvsuQyRH4CoTu4MgFjF"}]}]}} {"custom_id": "GF2_PMS2__L1A0001035795-MSS2_y04617_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8aZHNXSppKE8NgnHpAYG8q"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757317-MSS2_y03591_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2j8iqvpHa7HByDC8Lv5CFZ"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y02052_x05130.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-UX5Qm1xRDqW3CxYW4r8nHg"}]}]}} {"custom_id": "GF2_PMS2__L1A0000564692-MSS2_y05130_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-2e61jHj3fNB37pfJN9RDst"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y02565_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-APt2LMnzG1jyu8Mu56jiTU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001647941-MSS1_y06669_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-R8jCy6RF13JU8xyLDMHiMr"}]}]}} {"custom_id": "GF2_PMS2__L1A0001757320-MSS2_y02565_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SFmiDtJewQLc1YeTxQD9xR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001574001-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Twm4e94WVjkRLmxKhEFwaX"}]}]}} {"custom_id": "GF2_PMS2__L1A0001563129-MSS2_y00000_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-MXGjJZS3GBLEvKER1KzKax"}]}]}} {"custom_id": "GF2_PMS2__L1A0001119060-MSS2_y01539_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3BkPvtHmVgPsHd52nR1NHm"}]}]}} {"custom_id": "GF2_PMS2__L1A0001367840-MSS2_y02052_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-YB71munEPQtN7j6W2SmnHp"}]}]}} {"custom_id": "GF2_PMS2__L1A0001015594-MSS2_y01539_x02052.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-XKJvsSoziF9L7tH5x6mmkR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001787577-MSS2_y01026_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-FkiBE7VoEfrAnMjWjhFmh2"}]}]}} {"custom_id": "GF2_PMS2__L1A0001465929-MSS2_y04104_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-EPKJL4dEFvJgVrhtYr8bSW"}]}]}} {"custom_id": "GF2_PMS1__L1A0001366278-MSS1_y00513_x07182.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-H5YkgNqDRT89ZehCPFk6GU"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765572-MSS1_y04104_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-QvUBMXMdzzRjBSFqhmRVnf"}]}]}} {"custom_id": "GF2_PMS2__L1A0001642620-MSS2_y02052_x00000.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BxENc9aUsaGB5fKJRbf6py"}]}]}} {"custom_id": "GF2_PMS2__L1A0001206072-MSS2_y03078_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-7tPEfid1Hve97ktHKT5zr2"}]}]}} {"custom_id": "GF2_PMS1__L1A0001670888-MSS1_y00000_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RoipRr78G96rFPXJQjLKsA"}]}]}} {"custom_id": "GF2_PMS2__L1A0001396034-MSS2_y05643_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-9QRsZMZjYcyCbQFVoKqX8y"}]}]}} {"custom_id": "GF2_PMS1__L1A0001064469-MSS1_y05130_x03078.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-8ssDU8RXuTqSC8xuXAPjAF"}]}]}} {"custom_id": "GF2_PMS1__L1A0001015646-MSS1_y03078_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4jGW2T7A4teSZGbTFt4tEA"}]}]}} {"custom_id": "GF2_PMS1__L1A0001668483-MSS1_y05643_x05643.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BF4hhwYzmkxgMDqjhVqSca"}]}]}} {"custom_id": "GF2_PMS2__L1A0001566653-MSS2_y02052_x06669.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-X2p58k41EbgTcF8YqsWjPR"}]}]}} {"custom_id": "GF2_PMS2__L1A0001167480-MSS2_y03591_x04104.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-RkbLJRtMtTcaX7vN1p6ETw"}]}]}} {"custom_id": "GF2_PMS1__L1A0000564539-MSS1_y02565_x04617.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-Vt6rzbJJqoRh7s3PY9KpAP"}]}]}} {"custom_id": "GF2_PMS2__L1A0001490885-MSS2_y03591_x00513.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-BCkopezAEL8e1DUazCLUEs"}]}]}} {"custom_id": "GF2_PMS2__L1A0001116443-MSS2_y02052_x02565.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-3QwfUn61mcodqCeqNXX7gu"}]}]}} {"custom_id": "GF2_PMS1__L1A0000708367-MSS1_y06156_x01026.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-SMHvUurvtWkDb9Kr1YpiVT"}]}]}} {"custom_id": "GF2_PMS1__L1A0001680858-MSS1_y04104_x06156.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-LTf44gCCjFHA7xqYnQ4sNt"}]}]}} {"custom_id": "GF2_PMS1__L1A0001765574-MSS1_y03591_x01539.png", "method": "POST", "url": "/v1/responses", "body": {"model": "gpt-5.2", "input": [{"role": "user", "content": [{"type": "input_text", "text": "You are analyzing a satellite/aerial image.\n\nImage dimensions:\n- Width: 513 pixels\n- Height: 513 pixels\n\nYou are analyzing a satellite/aerial image. Examine the image carefully and identify all regions that match these land cover types: farmland, built-up, meadow, forest, and water.\n\nFor each land cover type, provide bounding box coordinates for EVERY distinct region you can identify in the image.\n\nCRITICAL: You must respond with ONLY a JSON object. Do not include any explanatory text, markdown formatting, or code blocks. Your entire response should be valid JSON that can be parsed directly.\n\nReturn this exact JSON structure:\n{\n \"built-up\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"farmland\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"forest\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"meadow\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...],\n \"water\": [[x1, y1, x2, y2], [x1, y1, x2, y2], ...]\n}\n\nCoordinate format:\n- x1, y1: top-left corner (percentages 0-100)\n- x2, y2: bottom-right corner (percentages 0-100)\n- Use empty array [] if a class is not visible\n- Include multiple boxes for the same class if there are multiple regions\n\nExample valid response:\n{\"built-up\": [[60, 10, 90, 40]], \"farmland\": [[10, 20, 50, 60]], \"forest\": [[5, 70, 30, 95]], \"meadow\": [], \"water\": [[70, 50, 95, 80]]}\n\nNow analyze the provided image and return only the JSON object with bounding boxes.\n"}, {"type": "input_image", "file_id": "file-4fsUp1ZB22nwpyqU7e1Nd8"}]}]}}